| name | reflect-skills |
| description | Use when the user wants to extract reusable skills from their AI session history and distribute them to their coding agents. |
reflect-skills
Run reflect skills to analyze your recent AI agent sessions and extract reusable skill patterns using an agent CLI.
Usage
reflect skills
reflect skills --all
reflect skills --week
reflect skills --agent gemini
reflect skills --agent claude
reflect skills --yes
reflect skills --demo
What it does
- Loads your local telemetry (same data as
reflect)
- Builds a deterministic evidence bundle from session quality scores, span/log-derived workflow signals, prompt snippets, recovery chains, and selected deep session context
- Invokes the specified agent CLI with an evidence-driven extraction prompt
- Shows you the extracted skills for review
- Asks for confirmation before writing
- Distributes confirmed skills to all detected agent directories (
~/.claude/skills/, ~/.cursor/skills/, etc.)
Extracted skills are immediately available via /skill-name in Claude Code.
Options
| Flag | Default | Description |
|---|
--agent | auto-detect | Agent CLI binary to use for extraction (prefers claude, falls back to gemini, codex, qwen) |
--yes / -y | off | Skip confirmation prompt |
--all / --week / --month / --day | --week | Time range for session analysis |
--demo | off | Use bundled sample data |
Example output
Running claude --print ...
Extracted 3 skills:
debug-loop Iterative debugging workflow with focused tool sequences
context-reset Pattern for clearing context and re-establishing scope
test-first-fix Test-driven bug fixing approach
Write these 3 skill(s) to 2 detected agent(s)? [Y/n]: y
✓ Claude Code: ~/.claude/skills/
✓ Cursor: ~/.cursor/skills/
3 skills ready. Use /debug-loop, /context-reset, /test-first-fix in Claude Code.
