Setup & Installation
openclaw skills install @stevengonsalvez/agent-reflect
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
npx clawhub install agent-reflect
What This Skill Does
Analyzes a conversation for correction signals and successful patterns, then proposes targeted edits to agent definition files and CLAUDE.md. Each accepted change is encoded permanently so the same correction doesn't need to happen twice.
Rather than re-stating preferences every session, corrections are written directly into agent files as diffs, making improvements persistent and auditable.
When to use it
- Encoding a TypeScript style correction into the frontend agent after a session
- Capturing a preferred commit workflow discovered during a project
- Selectively applying only certain changes from a batch of session learnings
- Auto-reflecting at session boundaries before context compaction
- Promoting a non-obvious debugging workaround into a reusable skill file
Example Workflow
Here's how your AI assistant might use this skill in practice.
INPUT
User asks: reflect
AGENT
- 1Scans the conversation for correction signals, categorized by confidence (HIGH: 'never', 'always', 'wrong'; MEDIUM: 'perfect', 'exactly')
- 2Maps each signal to a target file such as a named agent definition, CLAUDE.md, or a new skill file
- 3Checks whether any learning meets skill-worthy criteria: non-obvious, reusable, verified, and not already documented
- 4Presents a structured diff proposal with confidence levels and source quotes for each proposed change
- 5Applies approved changes to target files and updates metrics in ~/.reflect/
OUTPUT
Agent definition files updated with new rules; reflection log saved to .claude/reflections/YYYY-MM-DD_HH-MM-SS.md; metrics incremented in ~/.reflect/reflect-metrics.yaml