selftune

# selftune

> Self-improving skills for AI agents. selftune watches your real sessions,
> learns how you actually work, and evolves skill descriptions to match.
> Automatically. Works with Claude Code, Codex, OpenCode, and OpenClaw — no
> extra API keys, uses your existing subscription. Zero runtime dependencies.
> MIT licensed.

Your skills don't understand how you talk. selftune closes that gap — it
observes real sessions, detects missed triggers, and rewrites skill descriptions
to match how you actually work. A continuous loop: observe, detect, evolve,
validate, deploy, watch, repeat.

## Getting Started

- [README](https://github.com/selftune-dev/selftune/blob/HEAD/README.md): Project overview, quick start, and setup instructions
- [CONTRIBUTING](https://github.com/selftune-dev/selftune/blob/HEAD/CONTRIBUTING.md): Development setup, architecture rules, and PR guidelines
- [CHANGELOG](https://github.com/selftune-dev/selftune/blob/HEAD/CHANGELOG.md): Version history and release notes

## Architecture

- [ARCHITECTURE](https://github.com/selftune-dev/selftune/blob/HEAD/ARCHITECTURE.md): Domain map, module rules, and dependency structure
- [Golden Principles](https://github.com/selftune-dev/selftune/blob/HEAD/docs/golden-principles.md): Core design principles and invariants
- [Evolution Pipeline](https://github.com/selftune-dev/selftune/blob/HEAD/docs/design-docs/evolution-pipeline.md): How skill evolution works end-to-end
- [Monitoring Pipeline](https://github.com/selftune-dev/selftune/blob/HEAD/docs/design-docs/monitoring-pipeline.md): Post-deploy regression detection design

## Skill Integration

- [SKILL.md](https://github.com/selftune-dev/selftune/blob/HEAD/skill/SKILL.md): Agent-facing routing table for all selftune commands
- [AGENTS.md](https://github.com/selftune-dev/selftune/blob/HEAD/AGENTS.md): Agent configuration and behavioral guidelines

## Workflows

- [Initialize](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Initialize.md): Agent environment detection and config bootstrap
- [Evals](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Evals.md): Generate eval sets from real usage logs
- [Grade](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Grade.md): 3-tier session grading (trigger, process, quality)
- [Evolve](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Evolve.md): Propose, validate, and deploy improved skill descriptions
- [Watch](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Watch.md): Post-deploy monitoring and auto-rollback
- [Rollback](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Rollback.md): Restore pre-evolution skill descriptions
- [Replay](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Replay.md): Backfill logs from existing Claude Code transcripts
- [Contribute](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Contribute.md): Export anonymized data for community signal pooling
- [Doctor](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Doctor.md): Health checks on logs, hooks, config, and schema
- [Ingest](https://github.com/selftune-dev/selftune/blob/HEAD/skill/workflows/Ingest.md): Codex and OpenCode session ingestion

## References

- [Log Schema](https://github.com/selftune-dev/selftune/blob/HEAD/skill/references/logs.md): JSONL log file formats and record types
- [Grading Methodology](https://github.com/selftune-dev/selftune/blob/HEAD/skill/references/grading-methodology.md): How sessions are evaluated
- [Invocation Taxonomy](https://github.com/selftune-dev/selftune/blob/HEAD/skill/references/invocation-taxonomy.md): Explicit, implicit, contextual, and negative trigger types

## Optional

- [PRD](https://github.com/selftune-dev/selftune/blob/HEAD/PRD.md): Product requirements document
- [SECURITY](https://github.com/selftune-dev/selftune/blob/HEAD/SECURITY.md): Vulnerability reporting policy
- [CODE_OF_CONDUCT](https://github.com/selftune-dev/selftune/blob/HEAD/CODE_OF_CONDUCT.md): Contributor Covenant v2.1
- [Escalation Policy](https://github.com/selftune-dev/selftune/blob/HEAD/docs/escalation-policy.md): How regressions are triaged
- [Design Docs Index](https://github.com/selftune-dev/selftune/blob/HEAD/docs/design-docs/index.md): Index of all design documents