Agent Introspection

AI agent 自身の failure (loop / drift / max tool call / 環境ズレ) に対する構造化 self-debug。 capture → diagnose → contained recovery → introspection report の 4 phase で、 retry blind を防ぎ human escalation の前に agent が自己修正する。 Use when this capability is needed.

tomevault-io 229616d 2 files · 12.7 KB Updated

File contents

tomevault-io/skills-registry/tree/main/chronista-club--claude-plugin-chronista-style--agent-introspection commit 229616d581

Frequently asked questions

npx skillmds@latest add tomevault-io/agent-introspection