wall-of-apps-submission Developer Tools SignedAuthor: not available Submits or updates a Wall of Apps entry in the App-Store-Connect-CLI repo using the `asc apps wall submit` CLI, with a dry-run preview and a PR-based flow. Use when the user says 'submit to wall of apps', 'wall-of-apps', or 'add my app to the wall', wants to update an existing entry, asks for the exact Wall submission flow, or is submitting a pre-release or TestFlight app via a manual link, even if they do not name the CLI. View wall-of-apps-submission cmux-custom-sidebars Developer Tools SignedAuthor: not available Author, inspect, and verify cmux custom sidebar views written as interpreted SwiftUI-style .swift files under ~/.config/cmux/sidebars/. Covers the supported view subset, cmux sidebar validate/open/reload, confirming surface_type customSidebar, and debugging SwiftView interpreter render failures. Use when the user wants to build or edit a cmux custom sidebar, mentions interpreted Swift sidebars, the left sidebar picker, a sidebar that fails to render, or extending the sidebar data context — even if they never name the interpreter. View cmux-custom-sidebars lobe-ai-icons Developer Tools SignedAuthor: not available Build CDN URLs for and download AI/LLM brand icons — Claude, GPT, Gemini, Llama, DeepSeek, OpenAI, Anthropic, Hugging Face and 100+ more — from the lobe-icons library, or install its npm packages for React. Use when the user wants a logo or icon for any AI model, provider or app, in SVG/PNG/WEBP or light/dark/color variants, or asks in Chinese (e.g. 智谱, 月之暗面) for AI company icons, even if they never mention lobe-icons. View lobe-ai-icons opencli-adapter-repair Developer Tools SignedAuthor: not available Repairs broken opencli website adapters: captures a diagnostic RepairContext dump, explores the live page for new selectors or JSON APIs, patches only the adapter source, and retries up to 3 rounds. Use when an opencli command fails with SELECTOR, EMPTY_RESULT, API_ERROR, PAGE_CHANGED, TIMEOUT or COMMAND_EXEC errors, a site's DOM or API changed and its adapter broke, or the user wants the failing adapter fixed, not just reported. View opencli-adapter-repair contender-bakeoff Developer Tools SignedAuthor: not available Stages a parallel contest for a non-trivial artifact — several subagent contenders build independent drafts from one shared brief, an independent read-only judge scores them against a hidden rubric alongside the coordinator's own scoring, one draft becomes the base, and the strongest ideas from the rest are folded in by hand before the result is verified. Model lineup comes from a small local config file with within-vendor fallbacks. Use when the user asks for a bake-off, competing drafts, or a tournament of attempts, or when a single attempt at a complex deliverable would lock in one unexamined design. Not for trivial edits, quick lookups, or tasks cheap to redo.
View contender-bakeoff audit-npm-dependencies Developer Tools SignedAuthor: not available Read-only review of npm dependency health in JavaScript and TypeScript projects. Detects unused and misfiled packages, imports with no declared dependency, dead exports and orphaned files, libraries duplicating one role, tree-shaking misses, published advisories, and stale versions, using analyzers already installed in the project plus grep and glob fallbacks, grades each verified finding by severity, and checks Vitest configuration deprecations and leftover Jest artifacts when Vitest is present. Use when the user asks to audit, prune, or review dependencies, investigate bundle bloat, or question specific package.json entries. Not for installing, updating, or removing packages, applying security fixes, non-JavaScript ecosystems, or dictating the report's layout.
View audit-npm-dependencies biome-lint-and-format Developer Tools SignedAuthor: not available Sets up and runs Biome, a single native binary that formats and lints JavaScript, TypeScript, JSX, JSON, and CSS from one biome.json file in place of Prettier and ESLint. Covers fresh adoption, read-only audits, staged autofixes, importing Prettier and ESLint settings with a coverage checklist for unmapped rules, editor integration, and CI gating. Use when the user wants to adopt or configure Biome, move a project off Prettier or ESLint, apply autofixes safely, or centralize checks in scripts and pipelines. Not for bundling, type-checking, testing, runtime debugging, or languages Biome does not parse such as Python or Go.
View biome-lint-and-format codebase-example-scout Developer Tools SignedAuthor: not available Finds the strongest existing implementations of a requested feature or design pattern in a repository and distils them into an implementation guide with file paths, line numbers, conventions, pitfalls and test patterns. Queries the sourceatlas pattern lookup first and falls back to built-in code search only when that tool is unavailable or returns nothing usable. Use when the user asks how to implement something the way this project already does it, where a feature lives, or wants working examples before writing similar code. Not for designing features with no precedent in the repo, general code review, bug hunting or security audits.
View codebase-example-scout git-workflow-standards Developer Tools SignedAuthor: not available Guides git collaboration from end to end, covering choice of branching model, branch naming, Conventional Commits messages, pull request drafting and review, semantic versioning with annotated release tags, and changelog upkeep. Use when the user wants a commit message written or improved, a branch named, a release versioned or tagged, a changelog entry composed, a pull request drafted or reviewed, or team git standards settled. Not for resolving merge conflicts, recovering lost commits, debugging git itself, or rewriting history others have already fetched.
View git-workflow-standards pre-pull-request-audit Developer Tools SignedAuthor: not available Audits a finished feature branch against a base branch before a pull request is opened. Sizes the diff, walks the directory chain above each changed file for CLAUDE.md guideline files, then checks the patch for build and test breakers, guideline breaches, security flaws such as leaked credentials and injection, runtime bugs, and commit-history problems, using either one sequential pass or four parallel subagents. Produces a severity- and confidence-ranked report with a fix for every finding, plus optional verification of high-severity findings, optional execution of the project lint, test and build commands after user confirmation, and optional report-file output. Use when the user asks to scan, audit, review or check their branch, diff or changes before opening a PR or requesting review. Not for reviewing pull requests that are already open, running tests or linters unprompted, or applying fixes.
View pre-pull-request-audit recover-lost-claude-sessions Developer Tools SignedAuthor: not available Diagnoses Claude Code session-storage failures and guides their repair: an empty resume list, session jsonl files absent from the projects tree, files written under a temporary directory because HOME points somewhere else, an unwritable or missing projects folder, and older sessions stranded in a legacy layout awaiting consolidation. Use when resume reports no conversations, when session files are missing or appear in unexpected locations, when history vanishes after a machine, terminal, or IDE switch, or when legacy sessions must be gathered into the current tree. Not for errors inside a live conversation, account or billing trouble, or general questions about how Claude Code works.
View recover-lost-claude-sessions rust-smell-triage Developer Tools SignedAuthor: not available Guides Rust code review and incremental cleanup for recurring design smells: clone used to quiet the borrow checker, unwrap or expect on reachable failure paths, String and Vec parameters where a borrow or slice would do, hand-rolled index loops, unjustified unsafe blocks, and oversized functions. Surveys the code, groups findings by the design question each raises, lets the user choose the order, and lands one idiomatic fix at a time with compiler and test verification in between. Use when asked to review, audit, tidy, or modernize Rust code, or when clippy reports clone, unwrap, expect, needless_range_loop, or unused must-use diagnostics. Not for learning Rust from scratch, benchmark-driven performance work, or designing new unsafe FFI abstractions.
View rust-smell-triage wire-check-into-all-layers Developer Tools SignedAuthor: not available Installs one user-chosen code quality tool — linter, formatter, type checker, dead-code finder, or security scanner — into every enforcement layer of a uv-managed Python repository. Adds the dev dependency through uv, publishes a justfile recipe on the default target, registers a lefthook pre-commit job, and appends a GitHub Actions verify step, keeping configuration options identical across layers while deliberately varying file scope and check-only mode, and holding every file-rewriting behaviour behind explicit human approval. Use when the user names a specific check to add to just, lefthook, and GitHub Actions together. Not for choosing between tools, tuning rules, clearing reported violations, or projects on a different task runner, hook manager, or CI provider. View wire-check-into-all-layers rspack-v2-project-migration Developer Tools SignedAuthor: not available Moves an existing Rspack 1.x project to the 2.x line. Inventories every installed @rspack package with its exact version and dependency section, locates the config file, notes the package manager and Node version, reads the official v2 migration guide to determine which breaking changes apply, fills gaps in release tags, exact versions and plugin compatibility from official package metadata, presents an edit plan for user approval, upgrades each package inside its existing dependencies or devDependencies section, rewrites the config, and validates with build, dev-server, typecheck and test runs. Use when the user asks to upgrade or migrate a Rspack 1.x project to v2, or hits v2 breaking-change errors after bumping Rspack on their own. Not for webpack, Vite, Rollup, esbuild or tsup migrations, for Rsbuild or Rspress setups that pin their own Rspack version, or for adding Rspack to a brand-new project. View rspack-v2-project-migration structured-code-review Developer Tools SignedAuthor: not available Reviews pull requests and code diffs category by category, checking correctness, security, performance, maintainability, and test coverage, then reports ranked, actionable findings and a merge verdict. Use when the user asks to review code, inspect a PR or diff, evaluate code quality, or decide whether changes are ready to merge. Not for writing or refactoring the code under review, running builds or tests, or posting comments to hosting platforms.
View structured-code-review openspec-first-change-tour Developer Tools SignedAuthor: not available Teaches a newcomer the OpenSpec change workflow by coaching one small, real task in their own repository through a full cycle — readiness check, task selection, exploration, proposal, specs, design, task checklist, implementation, and archiving — with every artifact drafted in conversation and written to disk only after the user explicitly approves it. Use when a user is new to OpenSpec and asks for onboarding, a guided tutorial, a walkthrough, or help finishing a first change end to end. Not for experienced OpenSpec users who want a change created, continued, implemented, verified, or archived without the teaching narration, and not for projects where OpenSpec has not been initialized.
View openspec-first-change-tour personal-vscode-theme-builder Developer Tools SignedAuthor: not available Takes a color idea to a locally installed VS Code color theme by building a small extension from editable part files, merging them into one theme JSON, packaging a vsix with the vsce tool, and installing it through the code CLI so it appears in the theme picker. Use when the user wants a personal dark or light editor theme, wants syntax, workbench, or terminal colors restyled, or wants to update, repackage, or remove a theme built with this skill. Not for publishing to a marketplace, icon themes, general extension development, or editing VS Code settings files.
View personal-vscode-theme-builder plan-before-coding Developer Tools SignedAuthor: not available Turns a significant code change into a written implementation plan for the user to review and approve before any work starts, grounded in evidence read from the repository and organized around the decisions the user must make — approach, scope, completion criteria, and risk. Use when the user asks to plan a new feature, prepare a refactor or architectural shift, or settle an approach before implementation begins. Not for trivial fixes, quick questions about existing code, code review, or carrying out an already-approved plan.
View plan-before-coding plugin-generator Developer Tools SignedAuthor: not available Scaffolds a complete, installable Claude Code plugin from a user request: writes the manifest at .claude-plugin/plugin.json, creates only the component folders the purpose justifies, and fills each starter file in the format that component demands, including a hooks/hooks.json with the documented top-level hooks wrapper wired to a tested no-op starter script, and an .mcp.json only when a real server command exists. Adds a README, an optional local Git history, and a private test marketplace so the user can install and verify the scaffold before sharing it. Use when the user wants to create, bootstrap, package or bundle plugins, or prepare a team marketplace for distribution. Not for writing finished skill, agent, hook-script or MCP server logic, editing an existing plugin, or publishing to a remote host.
View plugin-generator abstraction-triage Developer Tools SignedAuthor: not available Helps decide between duplicating code and building a shared abstraction. Waits for a third genuine occurrence before extracting anything, spots helpers whose options each serve a single caller, walks through inlining a rotten helper back into its callers while preserving observable behavior — log strings included — and gates every code removal behind explicit user approval. Also weighs wrapper components and facades against their costs. Use when creating shared utilities or wrapper components, when tempted to add an option so a new caller fits an existing helper, or when refactoring an over-general abstraction. Not for style preferences, performance tuning, or system-level design such as service boundaries, data models, or API contracts.
View abstraction-triage finalize-openspec-change Developer Tools SignedAuthor: not available Closes a finished OpenSpec change once it is implemented, tested, approved and deployed. Checks closure preconditions, writes an archive summary with outcome, evidence, metrics, deviations, learnings and follow-ups, moves the change folder into a date-prefixed archive directory, merges the change's delta specs into the living specifications, validates the merged specs with strict checks, marks the proposal and task list as archived, optionally adds a changelog entry, and reports the closure result. Use when the user asks to archive, close, finalize, wrap up or mark complete an OpenSpec change, or says a deployed change is live and stable. Not for changes still in progress, failing tests, missing approvals, rolled-back releases, or partially deployed changes whose pending work has not been split out, and not for writing proposals or implementation code.
View finalize-openspec-change github-issue-lifecycle Developer Tools SignedAuthor: not available Runs the whole GitHub Issue life cycle through the gh CLI. Shapes bodies into bug, feature, epic, task, or brief formats, publishes single issues or batches after the user approves exact content, keeps labels, assignees, milestones, and bodies current, links issues as epic children or related work, records plan and branch to issue mappings in local JSON files, searches for existing issues before filing, and closes issues one by one or as a reviewed epic family. Use when the user asks to create, update, label, assign, link, find, sync, or close GitHub issues, or to file several related issues together. Not for pull requests, releases, GitHub Projects boards, or trackers other than GitHub Issues.
View github-issue-lifecycle primer Developer Tools SignedAuthor: not available Primes an agent for repository work through bounded, fail-closed orientation. It reads an append-only continuity ledger at .primer/STATE.md first, after verifying every path component, computes a verified git delta against the last recorded head, inspects a bounded set of manifests, README, and entry files, and reports purpose, stack, structure, entry points, dev loop with documented or observed required configuration names and prerequisites, current work, changes, open work, drift among inspected files, documented constraints, and up to five ranked next actions with signal, effort, and impact. It appends a strictly delimited JSON checkpoint to the ledger under a cooperative lock, preserving all prior bytes and owner notes. Offers Quick, Standard, and Deep depths with inheritance from the last full checkpoint. Writes only under user-authorized priming; otherwise operates read-only. Use when starting or resuming repository tasks, or when asked to prime, catch up, or decide what is next. View primer stripe-best-practices Developer Tools SignedAuthor: not available Guides Stripe integration decisions across development and test environment planning (separate sandboxes vs the shared test mode sandbox), API selection (Checkout Sessions vs PaymentIntents), Connect platform setup (Accounts v2, controller properties), billing/subscriptions, tax and registrations (Stripe Tax, automatic_tax, product tax codes), Treasury financial accounts, integration options (Checkout, Payment Element), migrating from deprecated Stripe APIs, and security best practices (API key management, API key permissions, webhooks, OAuth). Use when planning, building, modifying, testing, or reviewing any Stripe integration, including choosing a development environment, accepting payments, building marketplaces, integrating Stripe, processing payments, setting up subscriptions, collecting sales tax, VAT, or GST, creating connected accounts, or implementing secure key handling. View stripe-best-practices gh-pr-checks-proxy Developer Tools SignedAuthor: not available Bridges GitHub pull request checks and commit statuses through a local proxy when a fine-grained personal access token cannot read the Checks API directly. Fetches or watches check runs with the gh-checkproxy CLI, reads verdicts from exit codes, and hands every other repository action back to the normal gh CLI. Use when a fine-grained PAT makes PR CI status unreadable, when an automated workflow must wait on CI before proceeding, or when the proxy service needs configuring and starting on a trusted host. Not for writing check runs, for general repository operations, or for setups where the standard gh CLI already reads checks fine.
View gh-pr-checks-proxy incremental-mvp-builder Developer Tools SignedAuthor: not available Turns an existing plan document (commonly AGENTS.md) into a working MVP one feature at a time, using a propose, approve, build, verify loop with tests, lint and a commit after every feature, and keeps the plan roadmap and status notes current as work moves from scaffolding through release. Use when the user wants to start coding their MVP, implement or continue the next planned feature, or drive a phased build to completion. Not for authoring the plan, PRD or tech design docs themselves, for prototype spikes with no plan file, or for small fixes in a mature codebase.
View incremental-mvp-builder resolve-phpstan-errors Developer Tools SignedAuthor: not available Fixes PHPStan static analysis findings at their root cause by adding native type declarations, PHPDoc annotations, and small code changes, then re-runs analysis to confirm each error is gone. Covers missing parameter and return types, argument and return type mismatches, unknown properties and methods, array offsets, generics, dead code, and always-true or always-false comparisons, plus Laravel and Symfony specifics. Use when PHPStan reports errors, a CI or build step fails on PHPStan output, or the user asks for type hints, stricter typing, or static-analysis cleanup. Not for installing or configuring PHPStan from scratch, raising or lowering the analysis level, migrating from other analysers, or debugging PHP at runtime.
View resolve-phpstan-errors asc-identifier-lookup Developer Tools SignedAuthor: not available Resolves the numeric App Store Connect identifiers that asc commands require, covering apps, builds, app versions, in-app purchases, subscriptions and subscription groups together with their versions, localizations and images, TestFlight groups, testers and pre-release versions, and review submissions. Use when an asc command expects an ID but only a name, bundle ID, version string or build number is known, or when one lookup must be chained into the next. Not for creating, editing or deleting App Store Connect resources, and not for asc login or credential setup.
View asc-identifier-lookup distill-reusable-skills Developer Tools SignedAuthor: not available Turns teaching material such as tutorials, guides, and framework articles into a reusable executable skill, and reverse engineers finished works such as blog posts, READMEs, changelogs, and proposals to deduce the method behind them. Produces a skill with ordered steps, expected outcomes, verification lists, and pitfalls. Use when the user wants a document, web page, or sample work converted into a repeatable skill, or asks to learn from an example and produce similar output. Not for summarizing or explaining a document, answering questions about it, or converting file formats.
View distill-reusable-skills standalone-mise-tasks Developer Tools SignedAuthor: not available Teaches the agent to author mise file tasks, meaning standalone scripts placed in a mise tasks directory and annotated with MISE comment directives for description, dependencies, working directory and environment, plus USAGE comment directives that give the task flags, valued options and positional arguments, which mise injects as environment variables prefixed with usage_. Covers choosing between a file task and an inline mise.toml entry, reading the injected variables from bash, node or python, experimental monorepo task addressing, and verifying a task through task listings and generated help. Use when a task needs flags, options or positional arguments, carries logic too long for one line of mise.toml, must be written in a non-shell language, or must be selected per package inside a monorepo. Not for one-line tasks that fit inline in mise.toml, and not for installing or configuring mise itself.
View standalone-mise-tasks tmux-agent-status-triage Developer Tools SignedAuthor: not available Debugs broken AI-agent status detection in the tmux statusline plugin tmux-opencode-integrated. Covers listing agent panes, capturing pane text exactly as the detector reads it, reproducing verdicts with the plugin's libtmux-based smart_name helpers, repairing regex patterns on the ERROR, WAITING, BUSY, IDLE, UNKNOWN ladder, then proving the fix with pytest and rebuilding the package with Nix. Use when status icons mislabel agents such as opencode, claude or amp, when detection stops matching after an agent update, or when a new agent needs patterns. Not for sending keystrokes to agents, general tmux administration, or writing a statusline from scratch.
View tmux-agent-status-triage trace-execution-flow Developer Tools SignedAuthor: not available Turns questions about how a code path behaves into an auditable answer by running the flow tool with a natural-language query, checking the returned call graph against the repository, and presenting an ASCII tree annotated with API, DB, LIB and CLOUD boundary markers, entry points, and recursion or cycle findings. Covers whole-feature walks, single-function traces, failure branches, reverse caller searches and data-origin hunts. Use when the user asks how something works, what happens when an action fires, who calls a function, or where a value is set. Not for fixing bugs, writing new code, live debugging, or reading binaries without source.
View trace-execution-flow universal-bundler-plugin Developer Tools SignedAuthor: not available Author one compiler plugin and ship it as a single npm package that works across many build tools through unplugin — a shared typed factory, thin per-bundler adapter entries, subpath exports with optional peer dependencies, Nuxt and Astro wrappers, id-normalized transforms with accurate source maps, a playground with unit tests, packed-archive consumer verification in at least two bundlers, and read-only CI that runs those consumer builds. Use when creating, editing, reviewing, or verifying a cross-bundler build-plugin package, its adapters, exports, wrappers, types, tests, or CI. Not for configuring a Vite, Webpack, or Nuxt application that merely installs such plugins, and not for plugin logic that cannot share one factory. View universal-bundler-plugin issue-linked-branches Developer Tools SignedAuthor: not available Creates and maintains Git branches named after GitHub issues using the pattern issue-number, work-type, slug, for example 17-feature/infinite-scroll-for-long-result-tables. Derives names from issue titles and labels, renames nonconforming branches onto the pattern, validates names, keeps a branch-to-issue map at a fixed local path, switches to branches by issue number, and cleans up merged branches once their issues close. Use when the user wants issue-based branch names, needs a branch renamed to convention, asks to find, list, validate, or tidy issue branches, or wants a branch recorded on its issue. Not for commit messages, pull request review, release branching, or repositories that do not embed issue numbers in branch names.
View issue-linked-branches macos-bundle-stack-audit Developer Tools SignedAuthor: not available Identifies the technology stack of a macOS application bundle — Electron, Flutter, Qt, Mac Catalyst, SwiftUI, AppKit, or an undetermined native Apple stack — by delegating strictly read-only inspection of Info.plist, bundled frameworks, executables, linked libraries, code-signing records and Interface Builder artifacts to one command-running subagent per app, returning per-layer confidence ratings with quoted evidence lines. Use when the user supplies a .app path or an app name and asks what a Mac application is built with, whether it uses Electron, or which UI framework it uses. Not for launching or modifying applications, Windows or Linux binaries, iOS or Android packages, malware analysis, or decompilation.
View macos-bundle-stack-audit persistent-code-intelligence Developer Tools SignedAuthor: not available Guides an agent through building a persistent, locally stored understanding of one codebase with the in-memoria CLI and offering it to MCP clients such as Claude or Copilot. Covers deciding between durable memory and a one-off analysis, choosing exclusions, creating the store, validating it, keeping it current, serving it on a confirmed port, and repairing a stale or corrupt store. Use when the user wants agents to retain project knowledge across sessions, wants a codebase memory connected through MCP, or asks to refresh or fix an existing store. Not for single-use code summaries, projects small enough to read directly, installing the package, or machines where the CLI is absent.
View persistent-code-intelligence pipeline-trace-toolkit Developer Tools SignedAuthor: not available Read, query, and visualize JSONL execution traces from the traqo Python tracer, and add that tracing to Python programs. Covers raw and gzip trace files, zstd content sidecars, nested child traces, token and cost breakdowns for LLM calls, error triage with jq and the Python reader, the bundled web dashboard, conversion of Claude Code session transcripts into traces, and instrumentation via decorators, wrapped LLM clients, manual spans, and Claude Agent SDK hooks. Use when the user asks to inspect, debug, summarize, or chart a run saved as .jsonl or .jsonl.gz, dig into token spend, cache hits, slow steps, or failures, follow child trace files, turn Claude Code sessions into traces, or add tracing to Python code or LLM clients. Not for application logs that are not traqo traces, tracing frameworks in other languages, or changing the traced program's logic. View pipeline-trace-toolkit preserve-renamed-tool-names Developer Tools SignedAuthor: not available Guides renames of chat tool and tool set reference names so retired names keep resolving in saved prompt files, tool enablement state, auto-approval settings, and tool-set membership lists. Teaches where to record each retired name, which identifier fields are frozen, and how to review a diff that renames a tool. Use when renaming a toolReferenceName or a tool set referenceName, moving a tool between tool sets, or reviewing a pull request that edits tool registration code. Not for adding brand-new tools that had no earlier name, and not for renaming stable identifiers such as the id field or the package.json name field, which must never change.
View preserve-renamed-tool-names ray-local-http-debugging Developer Tools SignedAuthor: not available Posts JSON payloads over local HTTP to the Spatie Ray desktop app so values, tables, timing measurements, styled entries and rendered HTML previews appear live in the Ray window, with control over screens and windows. Use when the user asks to send, show, log, or preview something in Ray, to debug visually through Ray's own protocol, or to drive Ray screens and windows. Not for file or terminal logging, application logging frameworks, or debugging that does not involve the Ray app.
View ray-local-http-debugging ship-npm-package Developer Tools SignedAuthor: not available End-to-end guide for taking a JavaScript package from merged work to a public release — recording changesets, choosing semver bump levels, generating version bumps and the changelog, running release checks, dry-running and inspecting the tarball, publishing through npm or Bun, cutting a GitHub release with notes and assets, tagging, coordinating a monorepo, and verifying the result. Use when the user asks to version, changelog, publish, tag, or announce a package on the npm registry, or to record a changeset for finished work. Not for scaffolding new packages, CI or token automation setup, or ecosystems outside the npm registry.
View ship-npm-package stacked-branch-pr-workflow Developer Tools SignedAuthor: not available Guides the socle `so` CLI for chains of dependent Git branches and stacked pull requests, covering installing the binary, registering parent links, moving through a chain, replaying branches once their parents move, recovering from rebase conflicts, refreshing remote state, and opening one pull request per branch on GitHub. Use when the user drives stacked branches through the `so` command and wants to create, track, navigate, restack, sync, or submit a stack, or when so reports untracked branches or pauses on conflicts mid-restack. Not for single-branch Git chores, other stacking tools, GitHub work that bypasses `so`, or any push or history rewrite the user has not explicitly approved.
View stacked-branch-pr-workflow atomic-git-delivery Developer Tools SignedAuthor: not available Runs the Git delivery loop around small, single-purpose commits. Covers intent-named branches, conventional subjects that reference one tracker issue, a pre-commit quality gate with secret scanning, permission-gated Beads tracker sync, supervised auto-committing of finished tasks, and confirmation-gated push, pull request, merge, tagging and release notes. Use when the user asks to stage or commit changes, untangle mixed edits, link commits to tracker issues, automate commits for completed tasks, or prepare a release. Not for rewriting shared history or initializing repositories.
View atomic-git-delivery authoring-mage-targets Developer Tools SignedAuthor: not available Guides creating and maintaining Mage build scripts for Go repositories, where exported functions in files carrying the mage build tag become command-line targets. Covers the four valid target signatures and doc comments, dependency ordering through the mg package, running and capturing external commands with the sh helpers, timestamp-based up-to-date checks with the target package, namespaced target groups, a pinned tools sub-module using Go 1.24 tool directives, and confirmation gates around credential logins, file-rewriting lint fixes and publishing. Use when the user wants to write, extend, explain or debug a magefile or Go task automation such as build, test, lint, generate or clean targets. Not for Makefiles, Taskfiles, plain shell scripts, or CI provider configuration unrelated to Mage.
View authoring-mage-targets change-blast-radius Developer Tools SignedAuthor: not available Estimates the blast radius of editing a function, method, or class in the current repository by querying a preinstalled local GitNexus command-line program. Counts direct, transitive, and total dependents, lists the execution flows that pass through the symbol, optionally lists what the symbol itself depends on, finds the test files that exercise it, and grades the overall risk. Use when the user asks what breaks if a symbol changes, whether a symbol is safe to modify, or wants an impact check before a refactor. Not for making code changes, running tests, building or refreshing the code index, or analyzing code outside the indexed repository.
View change-blast-radius copilot-customization-workflow Developer Tools SignedAuthor: not available Guides the agent through choosing, placing, authoring, and debugging the markdown and JSON files that customize the VS Code Copilot agent: project-wide and path-scoped instruction files, prompt files, custom agent definitions, hook configurations, and packaged skills. Covers team-shared versus personal scope, frontmatter and description writing, tool restriction, hook lifecycle wiring with the input and output contract, and diagnosing customizations that are silently ignored. Use when the user wants Copilot to follow project or personal conventions, asks to scaffold, review, or repair one of these files, wants to limit an agent's tools, or needs shell commands to run at agent lifecycle moments. Not for general coding questions, runtime error diagnosis, MCP server setup, or VS Code extension development. Trivial fixes need only the fast path and final checks.
View copilot-customization-workflow deliberate-then-build Developer Tools SignedAuthor: not available Runs a structured deliberation pass before any non-trivial build. Agree with the user on the real need, gather several genuinely different strategies, research them without changing anything, score the survivors in an evidence matrix, and — on deep runs — build minimal throwaway prototypes of every finalist and record a measured comparison before choosing. Every depth ends with a written decision record the user accepts, and consequential actions during prototyping or implementation each get their own explicit confirmation. Use when a build or design request has several workable directions, or when the user asks which approach, stack, or architecture to pick. Not for questions with a single right answer, small fixes, formatting chores, or work already governed by an accepted record.
View deliberate-then-build design-doc-to-spec Developer Tools SignedAuthor: not available Rewrites a finished design document, such as a game design doc, product requirements doc, functional requirements doc or system design doc, into an implementation specification the user approves in stages, covering technology choices, data schemas, interface contracts, named constants, and small tasks carrying Given-When-Then acceptance criteria, a dependency graph and a git worktree workspace per task. Use when the user hands over a design document and asks for an implementation spec, a task breakdown, dependency ordering, decision records or per-task worktree setup. Not for writing the design document itself, reviewing code, or debugging a running system.
View design-doc-to-spec draft-pr-opener Developer Tools SignedAuthor: not available Prepares a draft GitHub pull request from the user's current feature branch. Reads the branch name and the diff against the base branch, derives a Conventional Commits style title, a structured Markdown description, a type-based label, and self-assignment, shows the exact content for approval, and only after an explicit yes pushes the branch and opens the PR in draft mode, then reports the new PR URL. Use when the user asks to open, create, prepare, or draft a pull request for the current branch, or wants consistent PR titles, descriptions, and labels generated from their changes. Not for merging, closing, reviewing, or marking pull requests ready, editing or committing code, installing or authenticating the GitHub CLI, or starting from protected branches such as main, master, or develop.
View draft-pr-opener gated-spec-workflow Developer Tools SignedAuthor: not available Runs a decision-gated workflow in which a written specification, an implementation plan, and a task list are each approved by the human before any code exists, a separate explicit authorization opens the coding phase, and the finished implementation is reviewed before it counts as done. Use when a new project or feature is starting, requirements are vague or only sketched, the change spans multiple files or modules, an architectural choice is pending, or the work will take more than 30 minutes. Not for typo fixes, single-line corrections, renames, or other self-contained changes whose outcome is already unambiguous.
View gated-spec-workflow cross-project-search-index Developer Tools SignedAuthor: not available Bundles several project directories into a single GrepAI workspace so one shared semantic index serves the whole group. Covers creating the workspace, registering each project, building and refreshing the combined index, querying everything or a single member, tracing a symbol across projects, reading status, adjusting membership, and wiring the workspace into an MCP client. Use when the user explicitly asks to set up, manage, index, search, inspect, modify, or remove a GrepAI workspace spanning multiple project directories, or to connect such a workspace to an MCP client. Not for searching inside one repository, for installing or upgrading GrepAI, for literal grep-style text lookups, or for conversations that merely mention monorepos or microservices without requesting workspace work.
View cross-project-search-index grepai-machine-output Developer Tools SignedAuthor: not available Selects the right machine-readable output shape for GrepAI searches and shows how to consume each one safely. Covers full JSON with matched code, compact JSON with shortened keys, TOON row output with its quoting and escape rules, jq, Python and Node.js consumers, saving results to files, opening hits in an editor, clipboard hand-off, and MCP tool calls with a format argument. Use when GrepAI results must feed a script, pipeline or AI agent context, or when a token budget forces the smallest workable payload. Not for interactive human-readable searching, building or refreshing the search index, or trace tool behaviour beyond its format options.
View grepai-machine-output live-docs-before-coding Developer Tools SignedAuthor: not available Pulls current reference material for third-party APIs, SDKs, and libraries through the locally installed chub CLI, so integration code follows today's interface instead of remembered and possibly outdated shapes. Saves short local notes per document id for future sessions and drafts documentation feedback for the user to approve. Use when the user asks to build against or call an external service, wants the newest API behavior, or mentions chub or Context Hub. Not for questions the project's own code can answer, machines where the chub command is unavailable, or live testing of endpoints.
View live-docs-before-coding maintainer-triage-desk Developer Tools SignedAuthor: not available Assists a project maintainer with tracker triage. Sorts each issue or outside pull request into one category and one status label, checks requests against the codebase for already-built behavior and past refusals, reproduces reported bugs locally, questions reporters whose requests are too thin to hand off, files refused feature ideas in a durable record so repeats are recognized, and writes handoff briefs an implementer can work from alone. Use when the user wants an attention queue over tracker items, a single item sorted or moved between states, or a repeat request checked against earlier refusals. Not for implementing the changes, merging pull requests, or planning releases.
View maintainer-triage-desk pinned-dotnet-cli-tools Developer Tools SignedAuthor: not available Maintains a repository-committed .NET tool manifest so every contributor and CI runner restores the same pinned builds of dotnet command-line utilities such as version-stamp calculators, coverage collectors, test-result converters, dependency freshness checkers, and license-gated static analyzers, all through one restore command. Covers creating the manifest, adding and hand-editing pinned entries, documenting one-time setup for newcomers, wiring restores into GitHub Actions and Azure Pipelines, refreshing and retiring pins, and fixing tool-resolution failures. Use when a repository must share dotnet CLI tools, when version drift between laptops or pipelines breaks builds, or when an existing dotnet-tools.json needs auditing, updating, or extending. Not for personal machine-wide tools, non-dotnet package ecosystems, or authoring and publishing a brand-new tool package.
View pinned-dotnet-cli-tools taze-upgrade-guardrails Developer Tools SignedAuthor: not available Refresh JavaScript and TypeScript dependencies by driving a locally installed taze CLI through a staged, confirmation-gated workflow that checks test coverage before risky bumps, filters which packages move, keeps standing policy in a config file, recurses across monorepo workspaces, previews every specifier change, and verifies results with tests, typecheck, and lint. Use when the user wants to update, bump, or upgrade npm, pnpm, yarn, or bun package versions, align workspace versions, or recover from an upgrade whose checks failed. Not for adding brand-new packages, publishing releases, security auditing, or ecosystems outside JavaScript.
View taze-upgrade-guardrails trace-outgoing-calls Developer Tools SignedAuthor: not available Uses the grepai trace callees command to list the outgoing calls it can detect statically inside one named function, with file, line, and source context for each call site, in text, full JSON, compact JSON, or TOON form. Use when the user asks what a function calls, what it depends on or touches such as storage, network, or logging, whether it has taken on too many responsibilities, or wants a dependency inventory for review or documentation. Not for discovering which functions call a target, for editing or refactoring code, or for installing and configuring grepai.
View trace-outgoing-calls doc-search-playbook Developer Tools SignedAuthor: not available Coaches the agent to run effective ImageKit documentation lookups through the search_docs tool: turning vague requests into precise queries, choosing among guides, community posts, API references, and SDK docs, and citing what comes back. Use when the user asks about ImageKit features, APIs, configuration, pricing, or SDK usage, when checking whether a capability exists, or when a transformation error needs parameter limits verified. Not for work unrelated to ImageKit, for performing uploads or edits itself, or as a replacement for official support channels.
View doc-search-playbook gated-git-workflow Developer Tools SignedAuthor: not available Runs a four-gate Git routine for coding agents — deciding what never enters history and cutting one branch per task, isolating parallel work in worktrees, verifying and inspecting each finished slice before an atomic, correctly typed commit, reporting every modification in three parts, diagnosing regressions with history tools, and merging, tidying up or publishing only after the user approves. Use when the agent edits files in a Git repository, stages or writes commits, opens branches or worktrees, juggles several tasks at once, or digs through history to explain a break. Not for non-Git version control, merge-conflict strategy, hosting-platform administration, or CI pipeline setup.
View gated-git-workflow voltagent-docs-search Developer Tools SignedAuthor: not available Looks up VoltAgent reference material inside the project's installed @voltagent/core package, where the markdown documentation travels with the exact dependency version. Locates the docs folder, lists its sections, greps for a keyword, and reads the matching files so answers match the installed release rather than the live website. Use when the user asks about VoltAgent APIs, agents, tools, workflows, guardrails, observability, deployment, or recipes, or when a version-specific signature matters. Not for questions about other libraries, general TypeScript help, or code that exists only in the user's own repository.
View voltagent-docs-search dotnet-assembly-internals Developer Tools SignedAuthor: not available Decompiles compiled .NET assemblies back into readable C# with the ILSpy command-line tool so the internals of DLLs can be studied. Use when a task requires understanding how a .NET API, NuGet dependency, framework library, or build output actually works and no source code is available. Not for assemblies whose license forbids reverse engineering, for defeating obfuscation, or for producing decompiled code that ships as-is.
View dotnet-assembly-internals retire-and-replace Developer Tools SignedAuthor: not available Walks the full retirement of a software system as a demolition project: a measured survey returning one of three verdicts (invest, build the replacement first, retire onto a production-proven successor), load diversion one caller at a time via traffic ramps, facades, or per-customer switches, and foundation-safe database changes staged as separate add, dual-write, backfill, read-switch, and remove releases, each reversible step shipped with a rehearsed down path and every individual production action carrying its own explicit human sign-off, including acknowledgment of one-way doors such as column removal. Use when consolidating overlapping implementations, retiring an API, library, or feature, moving consumers onto a successor, or reshaping a live production column without downtime. Not for incident response, greenfield builds, or general release and deployment mechanics.
View retire-and-replace ai-github-ops-playbook Developer Tools SignedAuthor: not available Teaches how to build GitHub Actions workflows that keep an AI model in the loop around repository work: machine-drafted pull request reviews, classification and labeling of newly opened issues, test selection driven by the changed paths, a risk assessment step in front of production deploys with a manual rollback path, comment-triggered rebase and cherry-pick helpers that stop for human confirmation, a mention bot with a fixed command set, and ownership plus branch-protection setup. Use when the user asks for a code review bot on pull requests, automatic issue triage, running only the test suites affected by a diff, an AI risk gate for deployments, or helper commands triggered by comments. Not for writing application source code, auditing an entire organization, or migrating CI from another provider.
View ai-github-ops-playbook everyday-bash-toolkit Developer Tools SignedAuthor: not available A working reference for Bash on Linux and macOS: joining commands with sequential, conditional and pipe operators, inspecting files and disk usage, reshaping text with grep, awk, cut, sort, uniq, sed and wc, examining processes and ports, setting environment variables, making curl requests, and building small scripts with strict error handling and exit traps. Use when the user asks to run, combine or debug shell commands, to write or harden a Bash script on a Linux or macOS machine, or to translate a PowerShell habit into Bash. Not for Windows-native PowerShell or CMD sessions, for writing programs in Python, Go or other languages, or for privileged system administration beyond everyday single-user tasks.
View everyday-bash-toolkit hive-config-coach Developer Tools SignedAuthor: not available Guides reading, writing and debugging the hive session manager's YAML configuration — per-repository rules, declarative tmux window layouts, Go template variables, palette commands, TUI keybindings, terminal spawn fallbacks and version migrations. Use when the user wants hive to open different windows or run setup for certain repositories, add custom commands or keys, integrate tmux or another terminal, or find out why a hive config edit is not taking effect. Not for tmux or terminal tweaks unrelated to hive, and not for hive's inter-agent messaging.
View hive-config-coach chrome-cli-automation Developer Tools SignedAuthor: not available Controls the user's real Chrome browser through a command-line tool for web testing and automation: page navigation, clicks and typing aimed at element refs or semantic roles, form filling, file uploads, screenshots, network and console inspection, device and throttling emulation, scripted multi-step runs, and questions sent to AI chat sites through existing logged-in sessions. Use when the user wants to automate or inspect a web page in Chrome, reproduce a UI bug, capture page screenshots, smoke-test a local site, or script a repeated browser task. Not for native desktop application control, pure API testing with no browser involved, or large-scale headless crawling.
View chrome-cli-automation discovery-sounding-board Developer Tools SignedAuthor: not available Runs a read-only discovery conversation alongside the user: clarifies fuzzy ideas, reads the codebase to ground the discussion, sketches ASCII diagrams, compares candidate approaches in tables, and offers, with consent, to record settled conclusions in the project's existing planning documents. Use when the user wants to reason through an idea, requirement, or design before committing to implementation, or is stuck partway through a planned change. Not for writing or editing application code, executing implementation tasks, or running a fixed step-by-step procedure.
View discovery-sounding-board explicit-problem-modeling Developer Tools SignedAuthor: not available Builds a written problem model covering data shapes, state, operations with preconditions and effects, invariants, and open questions, then freezes that model as a fixed agreement before any implementation code is written. Use when the user asks to model or specify before coding, or when a task involves state machines, business rules, constraint systems, or multi-step logic where correctness matters. Not for quick one-file scripts, simple text edits, or exploratory work with no rules to enforce.
View explicit-problem-modeling persistent-repo-notes Developer Tools SignedAuthor: not available Curates a repository-root AGENTS.md knowledge file so durable facts about a codebase survive across sessions. Guides reviewing recent work, proposing a numbered list of candidate facts, and merging only user-approved items into organized sections covering layout, commands, style, and workflows. Use when the user asks to remember, persist, or recall codebase knowledge, wants to create or update the repo memory file, or wants conventions and commands captured for future sessions. Not for one-off debugging notes, secrets or credentials, conversation history, or details relevant to a single issue.
View persistent-repo-notes ts-js-refactor-audit Developer Tools SignedAuthor: not available Audits TypeScript and JavaScript code for maintainability defects such as dead code, deep nesting, oversized functions, unsafe type assertions, barrel re-export files, and god modules, then reports findings grouped by category and applies minimal behavior-preserving fixes. Use when the user asks to clean up, simplify, tidy, or refactor TS or JS code, reduce complexity, or track down code smells. Not for implementing features, fixing bugs that need behavior changes, performance profiling, or languages other than TypeScript and JavaScript. View ts-js-refactor-audit world-clock Developer Tools SignedAuthor: not available Runs short inline Python snippets, using only the standard library zoneinfo database, to report the current time in any IANA time zone and to translate a wall-clock time from one zone to another, including daylight-saving status and the zone-to-zone gap for that date. Use when the user asks what time it is somewhere, needs a meeting, deadline, or call time moved between zones, or wants a time-zone name checked or found. Not for parsing log timestamps, calendar arithmetic such as adding weeks to a date, or stopwatch and timer features.
View world-clock behavior-safe-refactoring Developer Tools SignedAuthor: not available Refactors existing code to improve its structure while keeping observable behavior identical, verified by running the test suite after every small change. Covers mapping call sites, confirming behavior-focused test coverage before editing, matching code smells to specific fixes, ordering work from low to high risk, and applying React and TypeScript conventions. Use when asked to clean up, simplify, de-duplicate, rename, or reorganize code that already works. Not for adding features, fixing bugs, tuning performance without measurements, or rewriting modules from scratch.
View behavior-safe-refactoring commit-message-conventions Developer Tools SignedAuthor: not available Formats git commit messages in Conventional Commits style so changelog generators and semantic versioning tools can read the history reliably. Covers choosing the commit type, optional scope, lowercase imperative summary lines within length limits, bodies, footers, and breaking-change markers. Use when creating a git commit, drafting or rewording a commit message, checking a message against the convention, or when the user mentions commits, git history, or automated versioning. Not for choosing what to stage, writing changelogs or release notes, or formatting non-git text such as pull request descriptions. View commit-message-conventions draft-commit-messages Developer Tools SignedAuthor: not available Guides the agent through reading staged git changes and turning them into a well-structured commit message, covering conventional-commit type selection, subject and body drafting, ticket references, and handling of lock files, generated code, and oversized diffs. Use when the user asks to write, draft, review, refine, or create a commit for already-staged changes, or wants a diff summarized as a commit message. Not for pull-request descriptions, changelogs, release notes, or work that has not been staged yet. View draft-commit-messages edit-verify-confirm Developer Tools SignedAuthor: not available Runs coding work as small checked batches — read the surrounding code before editing, size the process to the job, run the cheapest useful automated check after every few edits, grow layered features from the core outward, and recover systematically when a fix stops working. Every finished batch is proposed to the user as a commit and waits for their explicit approval. Use when asked to write, build, implement, fix, or refactor code, especially across multiple files or steps. Not for deciding what to build, architecture debates, or read-only code review.
View edit-verify-confirm glab-gitlab-workflow Developer Tools SignedAuthor: not available Guides GitLab work through the glab command line tool: confirming the repository actually points at GitLab, creating branches and merge requests, watching pipelines, reading failed job logs to find root causes, retrying flaky jobs once, and handling issues and releases. Use when a git remote references a GitLab host and the user wants merge requests, pipeline debugging, issue triage, or releases handled with glab. Not for GitHub or other hosting platforms, for repositories with no GitLab remote, or for administering an entire GitLab instance.
View glab-gitlab-workflow guided-git-workflow Developer Tools SignedAuthor: not available Handles everyday Git tasks: reading repository status, staging files, writing Conventional Commit messages derived from the real diff, creating and switching branches, merging with step-by-step conflict help, stashing unfinished work, and pulling or pushing with a remote. Use when the user wants to commit changes, see what changed, create or switch branches, sync with a remote, stash work in progress, or resolve merge conflicts. Not for rewriting published history, force operations, editing repository configuration, or managing CI and release pipelines.
View guided-git-workflow install-github-skills Developer Tools SignedAuthor: not available Installs agent skills from GitHub repositories into the local skills directory and lists the skills a repository offers, marking ones already present. Fetches the repository archive for public repos or does a shallow sparse clone with the user's existing git credentials for private ones, verifies each skill folder has a SKILL.md file, and refuses to overwrite existing skills without confirmation. Use when the user wants to browse, list, or install skills from a GitHub owner/repo, a repo tree URL, or a skills collection. Not for authoring or editing skills, uninstalling skills, or fetching from non-GitHub sources.
View install-github-skills interactive-terminal-pilot Developer Tools SignedAuthor: not available Creates dedicated tmux sessions on a socket of its own and operates interactive terminal programs from the outside: launches detached sessions, types keystrokes into individual panes, reads the visible screen and its history, waits until a chosen line of output appears, and runs several such programs side by side. Use when a task needs a real interactive terminal, such as a conversational REPL, a wizard or installer that asks questions, a full-screen text interface, or many terminal programs to juggle in parallel. Not for work a plain background process already handles, and not for managing the tmux sessions the user already runs on their own server.
View interactive-terminal-pilot obsidian-app-control Developer Tools SignedAuthor: not available Runs vault operations against a live Obsidian desktop app through the obsidian CLI — reading notes, creating and appending content, daily-note journaling, vault-wide text search, frontmatter properties, and listings of tasks, tags and incoming links, with targeting by note name, vault-relative path or vault name. Also walks the reload-and-inspect verification loop for plugin and theme development, including in-app JavaScript evaluation, CSS inspection, screenshots, DOM queries and mobile emulation. Use when the user wants their Obsidian vault driven from the command line, asks for notes, properties or daily entries to be added or changed programmatically, or is iterating on an Obsidian plugin or theme and needs to check it inside the running app. Not for editing markdown files directly on disk while Obsidian is closed, and not for installing or configuring Obsidian itself.
View obsidian-app-control orphan-code-graph-sweep Developer Tools SignedAuthor: not available Finds functions no code calls, imports the resolver scored as low-confidence, and files with no import relationships by sending read-only Cypher queries to the codebase's GitNexus knowledge graph through the locally installed CLI, then weeds out common false positives and delivers a grouped report. Use when the user asks about unused or orphaned functions, dead code, broken or dangling imports, or isolated files, or wants to shrink surface area after a refactor or ahead of a release. Not for language-specific dead code linters such as Vulture or Knip, not for deleting the flagged code, and not when GitNexus is not installed and indexed on the machine.
View orphan-code-graph-sweep readable-code-refactoring Developer Tools SignedAuthor: not available Coaches code-quality review and behavior-preserving refactoring, covering intention-revealing names, small single-purpose functions, disciplined comments, error handling, unit-test hygiene, the SOLID design principles, and code-smell diagnosis with matching fixes. Use when reviewing or improving existing code, choosing or renaming identifiers, restructuring functions or classes, applying SOLID, hunting code smells, or writing unit tests. Not for choosing frameworks or libraries, configuring linters, formatters or CI, or performance tuning.
View readable-code-refactoring systematic-bug-diagnosis Developer Tools SignedAuthor: not available Guides a repeatable five-pass loop for fixing broken code: make the failure reproduce reliably, shrink the search space using the stack trace and data flow, name the underlying cause instead of the symptom, apply a minimal targeted repair, and prove it with type checks, tests, and the original reproduction. Includes playbooks for TypeScript compiler errors, runtime crashes, HTTP and API failures, React state bugs, performance bottlenecks, and module resolution problems, plus a hygiene checklist covering logging and regression tests. Use when the user reports an error, crash, failing build, failing test, or unexpected behavior, or asks why something stopped working. Not for writing new features, stylistic refactoring, or dedicated security audits.
View systematic-bug-diagnosis turbo-monorepo-tasks Developer Tools SignedAuthor: not available Guides configuration and daily use of Turborepo in JavaScript and TypeScript monorepos, covering package task definition in turbo.json, dependency wiring with dependsOn, output caching, environment variable hashing, filtering runs to changed packages, and CI integration. Use when the user edits turbo.json, debugs cache misses, structures apps and packages directories, shares code between apps, or wants to run only affected workspaces. Not for other build tools such as Nx, Lerna, or Bazel, or for publishing libraries to the public npm registry.
View turbo-monorepo-tasks turbo-workspace-orchestration Developer Tools SignedAuthor: not available Guides scaffolding and running Turborepo in JavaScript and TypeScript monorepos. Covers authoring turbo.json tasks with dependency ordering, cache inputs and outputs, per-package overrides, Next.js, NestJS and Vite task tuning, Vitest, Jest and Playwright pipelines, changed-package filters, and a GitHub Actions workflow that verifies only affected packages. Use when creating a Turborepo workspace, adding turbo to an existing repo, speeding up monorepo builds, or debugging task order and cache misses. Not for Nx, Lerna or Bazel, non-JavaScript projects, Docker image authoring, or production deployment setup.
View turbo-workspace-orchestration azure-devops-pr-flow Developer Tools SignedAuthor: not available Works with code hosted in Azure DevOps through its REST API and git: lists repositories, prepares branch-and-pull-request flows, opens draft PRs, and updates an existing PR's title or description. Use when the user mentions "Azure DevOps repo", "open a PR on Azure DevOps", "push my branch and create a pull request", "list our Azure DevOps repositories", or "update the PR description". Not for general Azure cloud administration, pipelines, Boards, or handling the user's tokens and remote-URL secrets. View azure-devops-pr-flow bitbucket-pr-workflow Developer Tools SignedAuthor: not available Drives a Bitbucket feature-branch and pull-request flow through the API — authenticating with the BITBUCKET_TOKEN environment variable, keeping PR identity stable across updates, choosing safe branches, and recovering from expired credentials without leaking them. Use when the user says "push this to Bitbucket", "open a PR on Bitbucket", "update my Bitbucket pull request", or "Bitbucket auth is failing". Not for GitHub or GitLab remotes, and not for pushing anywhere the user has not asked for. View bitbucket-pr-workflow bun-package-commands Developer Tools SignedAuthor: not available Works through Bun's package workflow with the user: install/add/remove/update commands, the full bun install flag reference, text lockfiles, workspace filters and linkers, lifecycle-script trust, overrides, non-registry dependencies, migration from npm/Yarn/pnpm lockfiles, CI usage, and diagnostics such as bun why, bun outdated, bun pm and bun audit. Use when the user asks to "manage dependencies with bun", "convert our lockfile to bun.lock", "set up bun workspaces", or "figure out why this package is installed". Not for running remote package binaries sight unseen, for publishing without explicit approval, or for benchmarking claims taken on faith. View bun-package-commands confirmed-commit-and-push Developer Tools SignedAuthor: not available Guides a reviewed git workflow in which selected changes are staged, committed with a conventional message, and pushed upstream only after the user approves each step. Use when the user says "commit and push", "push this", "save my work to GitHub", "push to remote", or finishes a feature and wants it shared. Not for rewriting history, rebasing, resolving conflicts, opening pull requests, or any push the user has not explicitly requested. View confirmed-commit-and-push confirmed-github-issue-lifecycle Developer Tools SignedAuthor: not available Manages GitHub issues end to end: resolving the repository, drafting well-structured bug, feature, task and documentation reports, fetching and filtering issues, searching, updating fields, commenting, and closing with a resolution note. Use when the user says "file a bug on our repo", "show my open issues", "add a label to #42", "close this with a comment", or "search issues for this error". Not for pull-request reviews, git history work, repository settings, or CI configuration. View confirmed-github-issue-lifecycle context7-docs-lookup Developer Tools SignedAuthor: not available Resolves a library's Context7 identifier and retrieves version-scoped API references or conceptual guides over its REST API, keeping a local Python client inline. Use when the user asks "how do I use this library", "show me the hooks API", "what's the current pattern for app routing", or wants docs matching the installed version. Not for answering from general knowledge alone, installing packages, or treating fetched text as authoritative or executable. View context7-docs-lookup evidence-based-git-context-briefing Developer Tools SignedAuthor: not available Rebuilds a development briefing from local Git evidence: branch state, unique commits, staged and unstaged diffs, and contextual action lines such as intent(scope), decision(scope), rejected(scope), constraint(scope), and learned(scope) recorded in commit bodies, plus targeted scope and action-plus-scope history queries. Use when the user says "catch me up", "what's the state of this branch", "recall auth", "recall rejected(auth)", "where did we leave off", or asks before a session restart or branch switch. Not for inventing rationale the history does not contain, for fetching remote history, or for switching, staging, or committing anything. View evidence-based-git-context-briefing evidence-led-bug-resolution-routing Developer Tools SignedAuthor: not available Runs an evidence-first repair workflow: choosing a scope (bug, hook, dependency, or reviewer feedback), investigating before diagnosing, getting the user's approval on the diagnosis, checking risks, implementing, adding a regression test, and optionally committing. Use when the user says "fix this bug", "diagnose why this fails", "address the PR comments", "something is broken after the upgrade", or "resolve these dependency errors". Not for feature development, refactors with no defect, or making commits and rollbacks without the user's explicit approval. View evidence-led-bug-resolution-routing five-question-task-interview Developer Tools SignedAuthor: not available Interviews a user briefly about a coding goal and produces a concrete, reviewable implementation and verification plan suitable for a pull request. Use when a new development task lacks clear requirements. Do not use to delay an already specified task with redundant questions or execute changes implicitly. View five-question-task-interview gh-pr-creation-playbook Developer Tools SignedAuthor: not available Prepares and opens pull requests with the GitHub CLI: checks branch readiness, reads the change against the remote base, writes a structured description with correct issue keywords, files post-merge follow-up issues, and sets draft, labels, reviewers, and base. Use when the user says "create a PR", "open a pull request", "submit for review", or "open a draft PR". Not for making commits or pushes on its own, and never for merging. View gh-pr-creation-playbook github-bug-report-writer Developer Tools SignedAuthor: not available Turns messy bug input — dictated ramblings, pasted error logs, screenshots, GIFs, half-remembered repro notes — into a crisp GitHub-flavored issue with numbered reproduction steps, environment facts, fenced error output, sanitized placeholders and a justified severity. Use when the user says "file a bug", "write this up as a GitHub issue", "turn my notes into an issue", or pastes an error and asks for a report. Not for publishing issues to GitHub without approval, for triaging an existing issue backlog, or for feature requests and discussions. View github-bug-report-writer github-pr-open Developer Tools SignedAuthor: not available Prepares concise pull request text from repository changes and opens an explicitly approved GitHub pull request with a branch push and self-assignment. Use when submitting completed branch work for review. Do not use for merging, changing repository access, or publishing unapproved commits. View github-pr-open git-worktree-isolation Developer Tools SignedAuthor: not available Manages isolated Git worktrees with explicit branch bases, status inspection, safe navigation, selective environment-file copying, and guarded cleanup. Use when the user asks to "review a PR in an isolated worktree", "work on two features in parallel", "switch between worktrees", or "clean up old worktrees". Not for discarding dirty work, silently copying secrets, or pushing, merging or deleting branches. View git-worktree-isolation ipynb-notebook-editing Developer Tools SignedAuthor: not available Reads, searches, edits and clears Jupyter notebooks as JSON, and runs or converts them with nbconvert. Use when the user asks to "fix a cell in this .ipynb", "clear all outputs", "find where this function is defined in the notebook", "execute the notebook", or "export it to HTML/script/Markdown". Not for running unreviewed notebook code, for building notebook-based dashboards, or for kernel and JupyterHub server administration. View ipynb-notebook-editing kaizen-code-improvement Developer Tools SignedAuthor: not available Organizes implementation, refactoring, validation, and process work into small verified changes, with error-proofing contracts, established team conventions, and just-in-time design, plus seven structured problem-analysis lenses. Use when the user asks to "refactor a messy module", "improve code quality incrementally", "stop this class of bug from recurring", "set up input validation", "cut over-engineering", or "run a root-cause analysis". Not for big-bang rewrites, speculative frameworks, unmeasured optimization, or making repository changes without approval. View kaizen-code-improvement mobile-web-debug-toolkit Developer Tools SignedAuthor: not available Diagnoses mobile web errors, touch behavior, viewport layout, and page timing using device inspectors, approved in-page consoles, automated capture, and privacy-aware telemetry on phones, tablets, and emulators. Use when the user says "debug this on my phone", "console errors on mobile", "test responsive layout on a real device", or "capture mobile page errors". Not for native app debugging, unapproved script injection, unrestricted production consoles, or claims that emulation equals real-device behavior. View mobile-web-debug-toolkit nx-monorepo-workflows Developer Tools SignedAuthor: not available Builds and operates Nx workspaces for TypeScript and JavaScript monorepos: creating projects, running generators, executing and caching tasks, computing affected sets, enforcing package boundaries, wiring React, Next.js and NestJS apps, setting up CI across major providers, and preparing releases. Use when the user says "create an Nx workspace", "generate a library", "run affected tests", "set up Nx CI", "configure Module Federation", or "speed up our monorepo builds". Not for publishing packages or pushing images without approval, for non-JavaScript build systems, or for Nx Cloud account administration. View nx-monorepo-workflows phased-task-execution Developer Tools SignedAuthor: not available Runs coding work through five gated phases — investigate, plan, implement, verify, complete — with evidence checkpoints before edits, real quality gates, and optional visual verification on web, desktop, mobile or CLI surfaces. Use when the user says "implement this", "fix this bug", "add this feature", "refactor this module", or gives any task that ends in changed code. Not for answering questions without code changes, running deployments, or committing and pushing without an explicit instruction. View phased-task-execution plugin-bundle-blueprint Developer Tools SignedAuthor: not available Designs and validates plugin bundles: manifest metadata, component directories, portable paths, hook documents, MCP server entries, plus a local scaffolding and structural-inspection toolkit. Use when the user asks to "scaffold a plugin", "write the plugin manifest", "organize commands and agents", "fix plugin discovery", "merge hook configs", or "check my plugin structure". Not for activating or publishing plugins, or for writing MCP servers themselves. View plugin-bundle-blueprint prepare-pr-head Developer Tools SignedAuthor: not available Prepares a reviewed pull-request head branch for merge: resolving every structured review finding, making scoped commits, running the project's gate sequence, and pushing with lease protection and verified handoff artifacts. Use when the user says "prepare the PR", "resolve review findings", "get the branch ready to merge", "push the review fixes", or asks to run gates before a merge. Not for opening reviews, merging or closing PRs, pushing to a protected base branch, or deleting worktrees. View prepare-pr-head production-readiness-audit Developer Tools SignedAuthor: not available Audits a codebase for production readiness across architecture, security, performance, code quality, testing and operations, then proposes and, with approval, applies prioritized repairs with measured before/after evidence. Use when the user says "make this production-ready", "audit my codebase", "find security issues", "check if we can ship this", or "optimize the code before launch". Not for deploying or publishing, creating tracker issues, or declaring compliance certifications. View production-readiness-audit single-call-pr-review Developer Tools SignedAuthor: not available Reviews a GitHub pull request and posts one bundled review through the API: prioritized inline comments on exact diff coordinates, multiline ranges, fenced suggestion blocks, and a brief summary body, assembled as a single JSON payload and submitted only with the user's confirmation. Use when the user asks to "review this PR", "post inline comments on a pull request", "leave review feedback on GitHub", "add suggestions to a PR review", or "approve or request changes on a PR". Not for reading PRs without posting, for resolving merge conflicts, or for editing the PR title and description. View single-call-pr-review topic-branch-creator Developer Tools SignedAuthor: not available Prepares validated git topic branch names (prefix and kebab-case subject), checks local and remote duplicates, and creates an approved branch — plain or linked to a tracker issue — with an optional upstream push. Use when the user says "create a branch", "start work on issue 412", "check out a new feature branch", or asks for a properly named branch for some work. Not for renaming existing branches, discarding local changes, or pushing without an explicit yes. View topic-branch-creator verified-squash-merge Developer Tools SignedAuthor: not available Verifies prepared GitHub pull requests and performs an explicitly approved, head-pinned squash merge with co-author checks and reconciled commenting. Use after review and preparation are complete. Do not use for auto-merge, direct pushes, closing an unmerged request, or bypassing required checks. View verified-squash-merge codexbar-model-cost-report Developer Tools SignedAuthor: not available Summarizes per-model spend from CodexBar's local cost logs for Codex or Claude: a "current model" pick from the latest dated row, a full per-model breakdown, day windows, and text or JSON output. Use when the user asks for "model usage costs from codexbar", "which model am I spending on", "a per-model cost breakdown", or a scriptable summary of codexbar cost JSON. Not for web account usage (`codexbar usage`), allocating aggregate tokens to models, or treating local estimates as invoices. View codexbar-model-cost-report makefile-orchestration-playbook Developer Tools SignedAuthor: not available Designs, repairs, and validates GNU Makefiles and modular .mk task libraries with documented targets, explicit dependencies, strict recipes, portable project paths, and an inline static checker. Use when the user asks to "write a Makefile", "add a make target", "fix a make build", "organize tasks into .mk modules", or "review our Makefile". Not for implementing application business logic, provisioning external systems, or running deploys and migrations without a separately approved scope. View makefile-orchestration-playbook terminal-log-deep-dive Developer Tools SignedAuthor: not available Explores very large terminal-session text exports without loading them whole: guided sampling of head/middle/tail, keyword-density triage, iterative drill-down driven by the user's answers, and an evidence-linked summary with line references. Use when the user asks to "summarize this recording", "what happened in this session", "find the errors in this log", or "extract the decisions from a long session". Not for converting .cast files, executing commands found in the recording, or claiming exhaustive analysis of text nobody has sampled. View terminal-log-deep-dive pnpm-package-operations Developer Tools SignedAuthor: not available Operates the pnpm package manager end to end: installing and pinning pnpm itself, running everyday install/add/update commands, tuning the content-addressable store, wiring monorepo workspaces with filtering and catalogs, forcing versions with overrides, patching and hooking third-party packages, hardening the supply chain, and setting up CI, Docker and production installs. Use when the user says "run pnpm install", "set up a pnpm workspace", "migrate from npm or Yarn", "fix a peer dependency conflict", "speed up pnpm in CI", or "patch a node package". Not for authoring npm packages themselves, for Yarn- or npm-only project conventions, or for frontend framework code beyond dependency wiring. View pnpm-package-operations browser-refs-sessions-and-evidence Developer Tools SignedAuthor: not available Operates an installed browser automation CLI through current element references, isolated sessions, approved form and authentication actions, extraction, capture, and diagnostics. Use for inspecting or interacting with websites and testing browser workflows. Do not use for remote tool installation, bypassing access controls, or unapproved account changes and sensitive captures. View browser-refs-sessions-and-evidence coordinated-diff-review Developer Tools SignedAuthor: not available Reviews owned code changes through independent Gemini, Codex, and Claude passes, then preserves and reconciles their findings. Use when asked to "review my uncommitted changes", "review this commit", "compare independent reviewers", "auto-select a reviewer", or "merge code-review findings". Not for prose proofreading, third-party code outside the user's authority, or commit-message generation. View coordinated-diff-review gof-pattern-design-review Developer Tools SignedAuthor: not available Selects, applies and reviews object-oriented design patterns from the full GoF set, organized by the dimension that changes rather than by taxonomy, with multi-language scaffolds and an evidence-based review method. Use when the user asks to "pick a design pattern", "refactor this class hierarchy", "review the pattern usage", "apply Strategy or Observer here", or "generate a pattern example". Not for adding ceremonial abstractions, unapproved code changes, or running examples with live side effects. View gof-pattern-design-review domain-first-code-design Developer Tools SignedAuthor: not available Steers everyday coding and design work toward domain-driven boundaries, reuse of proven libraries, and readable, focused modules. Use when the user asks to "design this feature", "review the architecture", "refactor this module", "decide between building or buying", or "clean up this code". Not for DevOps pipelines, infrastructure provisioning, or database performance tuning. View domain-first-code-design simplify-touched-code Developer Tools SignedAuthor: not available Tightens recently written or modified code for clarity and consistency while keeping its observable behavior exactly the same. Use when the user says "simplify this code", "clean up what you just wrote", "this function got hard to follow", or asks for a behavior-preserving refactor of the current change. Not for adding features, changing APIs or contracts, or reworking code outside the scope the user named. View simplify-touched-code mcporter-cli-bridge Developer Tools SignedAuthor: not available Drives any Model Context Protocol server from the shell with MCPorter: discovering configured servers, listing tools and schemas, calling tools over stdio or HTTP, managing persistent daemon connections, and generating TypeScript types or clients. Use when the user says "call this MCP tool", "what MCP servers do I have", "list tools on that server", or wants MCP capabilities without registering the server in their client. Not for building an MCP server from scratch, not for tools already registered in the host client, and not for running unreviewed downloaded server packages. View mcporter-cli-bridge git-feature-delivery-playbook Developer Tools SignedAuthor: not available Steers day-to-day git work from branch to review: picking a branching model, shaping atomic commits in conventional format, naming branches, opening clean pull requests, and undoing mistakes with amend, reset, revert, interactive rebase, cherry-pick, bisect and stash. Use when the user says "create a feature branch", "commit these changes", "open a pull request", "fix my last commit", "tidy my history before pushing", or asks for version-control best practices. Not for hosting-service administration, CI pipeline files, or migrating a repository to another VCS. View git-feature-delivery-playbook repository-work-start-verification Developer Tools SignedAuthor: not available Verifies a repository before implementation starts: refreshes remote state, looks for existing issues and pull requests covering the work, measures branch divergence, flags dirty files and stashes, and predicts merge conflicts with a dry-run. Use when the user says "start on issue 412", "pick up where I left off", "create a branch for this fix", "check before I begin", or is about to spawn parallel agents on one repo. Not for quick single-file edits, actually resolving conflicts, or performing the merge itself. View repository-work-start-verification tested-skill-upstream-contribution Developer Tools SignedAuthor: not available Prepares one tested, broadly useful skill for contribution to an upstream skills repository as its own branch and pull request: suitability check, fork sync, dedicated branch, commit, push, PR body, and post-merge cleanup. Use when the user says "contribute this skill upstream", "open a PR for my skill", "share this skill back to the repo", or "publish my local skill". Not for private or organization- specific skills, mass skills-database operations, or unrelated repo housekeeping. View tested-skill-upstream-contribution deep-module-refactor Developer Tools SignedAuthor: not available Walks an agent through collapsing a sprawl of thin, hard-to-test code into a few modules that hand callers a large amount of behaviour for a short list of learned facts. Covers a fact-counting audit before judging any code shallow, a four-way dependency triage that settles where swap points and adapters belong, testing that observes the module exactly as a caller would, diagnosis of tests broken by internal refactors, a human confirmation gate before deleting code or tests, and a parallel exploration pattern for weighing rival interface shapes. Use when the user wants to deepen or redesign a module's interface, place a seam, improve a cluster's testability, or compare candidate designs. Not for UI and visual work, performance tuning, build and CI tooling, dependency upgrades, or framework selection.
View deep-module-refactor verified-branch-integration-choice Developer Tools SignedAuthor: not available Walks a completed development branch through a verified handoff: running the test suite, then offering a confirmed choice between local merge, pull request, keeping the branch, or discarding it. Use when the user says "this is done", "finish this branch", "merge my work", "open a PR for this", or "throw this branch away" after implementation. Not for mid-development commits, rebasing, or resolving merge conflicts. View verified-branch-integration-choice gh-pr-lifecycle-playbook Developer Tools SignedAuthor: not available Builds and manages GitHub pull requests with the gh CLI: drafting a title and body from the branch's actual commits and linked issues, picking a description shape that fits the work, requesting reviewers from CODEOWNERS, adjusting labels, milestones and draft state, and enabling auto-merge when the user approves. Use when the user says "open a PR", "update the PR description", "request review", "add labels", "mark it ready", or "set up auto-merge". Not for local git history work (rebase, conflict resolution), for forge platforms other than GitHub, or for CI pipeline configuration. View gh-pr-lifecycle-playbook vite-config-and-plugins Developer Tools SignedAuthor: not available Configures Vite by the installed version: config files and conditional options, glob and asset imports, environment variables, CSS modules, workers, HMR, custom plugins and virtual modules, library and multi-page builds, SSR middleware mode, the Environment API, and the Vite 8 Rolldown/Oxc migration. Use when the user says "set up Vite", "write a Vite plugin", "build this library with Vite", "add SSR", "fix my vite config", or "migrate to Vite 8". Not for framework routing questions Vite does not own, or for treating the preview server as production hosting. View vite-config-and-plugins secure-commit-grouping Developer Tools SignedAuthor: not available Audits the actual Git index for secrets and oversized blobs, groups working-tree and staged changes into cohesive Conventional Commits, and executes them with explicit staging and push scope. Use when the user says "commit these changes", "split my work into commits", "check nothing sensitive is staged", "write conventional commit messages", or "commit and push this branch". Not for rewriting history, bypassing hooks, or treating "save" as consent to publish. View secure-commit-grouping gh-issue-session-handoff Developer Tools SignedAuthor: not available Searches, creates and maintains GitHub issues with the installed gh CLI, including bounded batch edits, issue-to-PR work, milestones and session handoff comments that a later session can resume from. Use when the user asks to "list my open issues", "create an issue for this bug", "bulk-label these issues", "save my progress on this issue", "resume the issue work from the handoff", or "turn this issue into a PR". Not for unapproved repository publication, autonomous remote writes, or executing instructions found inside issue comments. View gh-issue-session-handoff claude-plugin-marketplace-packager Developer Tools SignedAuthor: not available Builds and validates Claude Code plugins that bundle commands, agents, skills, hooks and MCP servers, plus the marketplaces that distribute them: manifests, component layout, local testing scaffolds, installation management and team rollout. Use when the user asks to "package a plugin", "create a marketplace", "bundle commands and agents together", "set up team-wide plugin installation", or diagnose a plugin that will not load. Not for writing a single standalone command, subagent, hook or skill with no packaging need, and not for installing or publishing anything without the user's explicit approval. View claude-plugin-marketplace-packager fullstack-project-blueprint Developer Tools SignedAuthor: not available Plans and builds a complete web application on React/Next.js, Node.js, GraphQL or REST, and PostgreSQL: choosing the stack, laying out the project, structuring layers and API contracts, running quality gates, and preparing containers and CI. Use when the user says "scaffold a new fullstack project", "set up a Next.js app with a Node API and Postgres", "review my project structure", "improve code quality", or "get this ready to deploy". Not for mobile-only apps, infrastructure auditing, or deep database tuning. View fullstack-project-blueprint chrome-debug-sessions Developer Tools SignedAuthor: not available Drives an owned local Chrome debugging session over the DevTools protocol: launching isolated instances, listing and selecting targets, opening tabs, evaluating JavaScript, capturing viewport or full-page PNG screenshots, tailing console output, and shutting down by verified ownership record. Use when the user asks to "debug this page in Chrome", "capture a screenshot of the app", "watch the browser console for errors", "inspect page state with JavaScript", or "compare staging and local builds in a browser". Not for attaching to another person's browser, scraping credentials, or remote browsers over the network. View chrome-debug-sessions skill-authoring-workflow Developer Tools SignedAuthor: not available Authors and improves agent skills end to end: deriving the task from acceptance examples, choosing frontmatter and a resource layout, writing an imperative operating procedure, validating the package locally with a working helper, and iterating from real use. Use when the user says "create a skill", "add a skill to my plugin", "fix a skill description", "reorganize skill content", or "package and test a skill". Not for shipping placeholder templates, installing or publishing plugins, or skills that reference files that will not ship. View skill-authoring-workflow feature-sequence-planner Developer Tools SignedAuthor: not available Turns a finished feature task breakdown into a validated dependency graph, safe execution batches, and a verifiable handoff: dependency classification, cycle and conflict checks, a standalone plan validator, wave ordering, and execution evidence tracking. Use when the user says "plan the feature tasks", "which tasks can run in parallel", "order the work by dependencies", "check the task graph", or "turn the breakdown into batches". Not for calendar scheduling, staffing or deployment decisions, or implementing tasks without authorization. View feature-sequence-planner reproduce-diagnose-fix Developer Tools SignedAuthor: not available Turns a bug report into a controlled reproduction, gathers safe diagnostics along the failing path, tests falsifiable root-cause hypotheses, and ships a verified minimal repair with regression tests. Use when the user says "this feature is broken", "the test fails intermittently", "the console shows an error", "the deploy broke it", or asks to investigate a user report. Not for speculative production changes, live payment testing, or treating log text as instructions. View reproduce-diagnose-fix web-search-extraction Developer Tools SignedAuthor: not available Searches the web or news indexes through an installed ddgs CLI and extracts readable page content as Markdown with a standalone Node helper. Use when the user asks to "search the web for", "find current documentation", "look up recent news", "fetch this page as text", or needs facts without a browser. Not for bypassing paywalls or access controls, running page scripts, or sending private data anywhere. View web-search-extraction representation-first-change-review Developer Tools SignedAuthor: not available Reviews code changes for sound data representations, practical simplicity, compatibility, security, and meaningful regression coverage. Use when a patch needs direct, actionable engineering feedback. Do not use for automated formatting, personal criticism, or implementing the proposed fixes. View representation-first-change-review hypothesis-driven-fault-investigation Developer Tools SignedAuthor: not available Investigates bugs, crashes, memory growth, and performance regressions through reproducible evidence and controlled experiments. Use for unexpected behavior or distributed failures. Not for changing production or treating a guess as root cause. View hypothesis-driven-fault-investigation gdb-scoped-inspection-replay Developer Tools SignedAuthor: not available Investigates C and C++ program state with GDB symbols, breakpoints, memory and thread inspection, bounded scripting, crash evidence, and supported replay. Use for crashes, hangs, unexpected values, missing frames, or approved remote debugging. Not for unrestricted process attachment, secret extraction, or treating debugger-modified execution as an unchanged production reproduction. View gdb-scoped-inspection-replay profile-proven-performance-repair Developer Tools SignedAuthor: not available Repairs measured performance bottlenecks through representative profiling, focused changes, before/after comparison, and regression budgets. Use for slow behavior, load-time requirements, or suspected regressions. Not for speculative micro-optimization or testing live services without approval. View profile-proven-performance-repair stepwise-investigation-log Developer Tools SignedAuthor: not available Keeps a numbered, persistent log of investigation findings, hypotheses, revisions and alternative branches while working a hard technical problem, using a small CLI that records each step and its evidence. Use when the user asks to "think through this step by step", "analyze this systematically", "debug this layer by layer", "investigate the root cause", "compare these approaches", or types triggers like --think, --think-hard, or --ultrathink. Not for simple one-step answers, single-file fixes, quick explanations, or requests to expose private model reasoning. View stepwise-investigation-log measured-app-optimization Developer Tools SignedAuthor: not available Optimizes an application across four measured dimensions — speed, dependencies, database, and dead code — by auditing with real numbers, proposing scoped fixes, and verifying each change against a baseline. Use when the user says the app is "slow", asks to "reduce bundle size", "clean up package.json", "remove unused code", or "speed up database queries". Not for building features, running production load tests, or unapproved deletes, installs, or schema migrations. View measured-app-optimization measured-application-latency-improvement Developer Tools SignedAuthor: not available Investigates frontend, API, database, and serverless bottlenecks with baselines, focused changes, and load/behavior validation. Use for slow pages or production-readiness performance reviews. Not for testing public services, installing profilers, or altering indexes without approval. View measured-application-latency-improvement staged-stripe-version-migration Developer Tools SignedAuthor: not available Plans Stripe API, server SDK, browser, and mobile upgrades through version inventory, product-specific compatibility checks, fixtures, and staged rollout. Use for an explicitly requested upgrade. Not for assuming the latest version, modifying account defaults, or calling live payment APIs automatically. View staged-stripe-version-migration completed-change-pr-brief Developer Tools SignedAuthor: not available Produces a concise pull request title, description, and branch suggestion from completed local work, including committed and untracked changes. Use when wrapping up a substantive change for review. Not for opening a PR, editing branches, or fabricating compatibility risks. View completed-change-pr-brief verified-pull-request-body-update Developer Tools SignedAuthor: not available Repairs pull request description updates that report success without persisting, using a body file, an approved REST fallback, and exact read-back verification. Use for multiline PR body changes or silent gh edit failures. Not for title-only edits, automatic remote writes, or guarantees that an API never fails. View verified-pull-request-body-update idea-to-buildable-spec Developer Tools SignedAuthor: not available Turns a product idea into either a compact feature specification an AI builder can implement, or a complete project scope with roles, screens, integrations, milestones and cost estimates for stakeholders and contractors. Use when the user says "write a spec for this feature", "define our MVP", "I need a project scope for a contractor quote", "AI keeps building the wrong thing", or asks to plan scope and budget before development. Not for writing the actual implementation code, managing projects after the spec exists, or detailed UX wireframing. View idea-to-buildable-spec scoped-github-cli-work Developer Tools SignedAuthor: not available Inspects and manages GitHub issues, pull requests, reviews, CI runs, and repository API data through gh with explicit target scope. Use for requested GitHub operations. Not for local Git work, non-GitHub hosting, complex browser flows, or unrestricted bulk changes. View scoped-github-cli-work prompts-catalog-skill-discovery Developer Tools SignedAuthor: not available Searches the prompts.chat skill catalog, retrieves complete packages, and installs a selected skill after explicit approval. Use when a user wants reusable agent capabilities or a named skill from that catalog. Not for publishing a package or executing its downloaded helpers. View prompts-catalog-skill-discovery coordinated-plugin-version-release Developer Tools SignedAuthor: not available Prepares a plugin release by aligning skill and extension versions, writing release notes, checking shared utilities, and coordinating reviewed commits, pull requests, tags, and release records. Use for an explicitly requested version release. Not for automatically pushing, merging, or publishing. View coordinated-plugin-version-release indexed-execution-walkthrough Developer Tools SignedAuthor: not available Explains repository behavior by tracing GitNexus execution flows and symbol relationships back to local source. Use for code walkthroughs, function responsibilities, or unfamiliar subsystems. Not for editing code, automatic reindexing, or claiming static analysis proves runtime behavior. View indexed-execution-walkthrough clawhub-package-lifecycle Developer Tools SignedAuthor: not available Uses ClawHub to locate, list, install, update, and publish agent skill packages with explicit version and destination choices. Use for managing a ClawHub skill installation or preparing a confirmed package release. Not for executing downloaded helper code or publishing to another catalog. View clawhub-package-lifecycle evidence-first-codebase-orientation Developer Tools SignedAuthor: not available Explains an unfamiliar repository through observed stack, directories, architecture, entry points, conventions, and development commands. Use for onboarding or architecture questions. Not for modifying code, executing project setup automatically, or guessing technologies from filenames alone. View evidence-first-codebase-orientation diff-led-conventional-commit-recording Developer Tools SignedAuthor: not available Derives a conventional commit message and logical staging proposal from the actual git diff, supports user overrides, and records an approved new commit. Use when a user asks to commit or save repository changes. Not for pushing, rewriting history, changing git configuration, or bypassing failed hooks automatically. View diff-led-conventional-commit-recording deno-project-toolkit Developer Tools SignedAuthor: not available Works in Deno projects: detecting the toolchain, scaffolding with deno init, adding and removing JSR and npm dependencies, running scripts with the smallest sufficient --allow-* permission set, executing defined tasks, and using the built-in fmt, lint, and test commands. Use when a repository contains "deno.json" or "deno.jsonc", uses "deno.lock", or the user asks to "add a Deno dependency", "run a deno task", or "deno fmt this project". Not for Node.js or Bun projects without a Deno config, and not for publishing packages to JSR or npm. View deno-project-toolkit cached-symbol-atlas Developer Tools SignedAuthor: not available Builds, updates, and audits a local AST-derived repository map of symbols and imports with ast-grep. Use for map creation, map status, code drift, or symbol inventory. Not for executing source code, installing parsers automatically, or proving runtime behavior from static structure. View cached-symbol-atlas pnpm-catalog-upgrade-and-recovery Developer Tools SignedAuthor: not available Plans and verifies incremental dependency upgrades using pnpm catalogs, compatibility review, security checks, and controlled recovery. Use for patches, major migrations, or dependency conflicts in a pnpm workspace. Do not use for resetting work or downloading and executing codemods without approval. View pnpm-catalog-upgrade-and-recovery local-skill-suite-onboarding Developer Tools SignedAuthor: not available Plans and verifies installation of a user-selected local behavioral skill suite through reviewed symlinks, with optional task and memory integrations. Use for a new machine or missing installed lifecycle skills. Not for overwriting custom links or automatically executing downloaded code. View local-skill-suite-onboarding Give your agent the ones it needs. Start free with five active skills for automatic loading and manual access to the whole library. Upgrade for automatic loading across the library.
Create a free key