agent-tool-approval-patterns

Designs the internals of autonomous coding agents: the bounded propose-approve-execute loop, tool schemas and granularity, permission levels and approval UX, real sandboxing, browser observation, context selection, checkpointing, and MCP tool adapters. Use when the user says "build a coding agent", "design a tool API for my agent", "add an approval flow before file edits", "sandbox the commands my agent runs", or "hook my agent up to MCP tools". Not for prompt-only chatbots with no tool use, for bypassing approval gates in an existing agent product, or as a substitute for OS-level isolation.

Version

v1.0.1

Size

16 KB

Updated

2026-10-06

Your client verifies the signature before reading. This explicit manual read is available on every plan. Free automatically loads five chosen skills; paid plans automatically load relevant skills across the library.