learning-capturelisted
Install: claude install-skill wagneripjr/skills
# Learning Capture
Log learnings, errors, and feature requests to structured markdown files in `.learnings/` so what one session learns reaches the next. High-value learnings get promoted to project memory (`CLAUDE.md`) or cross-project memory (auto memory).
**Two memory tiers.** `.learnings/` entries are *episodic* memory — specific, timestamped experiences. Promotion distills proven ones into *semantic* memory — durable rules in `CLAUDE.md`, auto memory, or skills. The experience stays logged; the lesson graduates to where every session reads it.
The entry grammar is a contract other tools read: [references/entry-format.md](references/entry-format.md). Worked entries: [references/examples.md](references/examples.md). Claude Code versus Codex: [references/host-runtime.md](references/host-runtime.md). `<skill-dir>` below is the directory holding this `SKILL.md`.
## Quick Reference
| Situation | Trigger Phrases | Action |
|-----------|----------------|--------|
| Command/operation fails | Non-zero exit, stack trace, timeout | Log to `.learnings/ERRORS.md` |
| User corrects you | "that's wrong", "actually...", "no, it should be..." | Log to `.learnings/LEARNINGS.md`, category `correction` |
| User wants missing feature | "can you also...", "I wish you could...", "is there a way to..." | Log to `.learnings/FEATURE_REQUESTS.md` |
| API/external tool fails | Connection refused, timeout, unexpected response | Log to `.learnings/ERRORS.md` with integration details |
| Knowledge