self-improving-agentlisted

# Self-Improving Agent Curate session memory, promote recurring patterns to permanent config, and extract reusable skills. ## Memory Stack Three tiers, from most permanent to most ephemeral: 1. **Repository guidance** (you write) -- AGENTS.md in Codex or CLAUDE.md in Claude Code. Project-wide rules, preferences, and conventions that survive across sessions. 2. **Review memory** -- The host memory store, or a project MEMORY.md when the user has chosen a file-backed ledger. Observations, patterns, and corrections awaiting review. 3. **Session Memory** -- In-context learnings that exist only for the current session. Lost on exit unless promoted. ## Commands | Command | Action | |---------|--------| | `review` | Scan the active review-memory source for promotion candidates. Show each with recurrence count and recommendation. | | `promote` | Promote a pattern to AGENTS.md, CLAUDE.md, or an appropriate scoped rule. Remove it from review memory after promotion. | | `extract` | Generate a complete skill from a recurring pattern (creates `skills/<name>/SKILL.md`). | | `status` | Show memory stats: entries, staleness, promotion candidates, and session observations. | | `remember` | Capture a specific observation with timestamp and context. | Invoke these as `$self-improving-agent <command>` in Codex, `/si:<command>` in Claude Code, or describe the operation in natural language. ## Promotion Lifecycle See [promotion-lifecycle.md](./promotion-lifecycle.md) for detailed rules. S