# ASI (Agent Skill Interface)

> ASI is a behavioral contract for agentic systems that defines determinism-before-reasoning, strict passivity, explicit failure semantics, and auditable scope.

---

## Just Use It

Skip the theory. If you're in an environment with ASI skills available, invoke directly:

```text
/asi-onboard   → (Optional, recommended) Build repo context (produces NOTES.md + SOURCES.md)
/asi-kickoff   → Design a new skill (produces KICKOFF.md)
/asi-plan      → Kickoff + plan (produces KICKOFF.md, then PLAN.md + TODO.md)
/asi-exec      → Execute with receipts (produces RECEIPT.md)
```

The skills enforce ASI principles automatically. You don't need to understand the spec to benefit from it.

---

## Understand It

Use the links below as the recommended progressive disclosure entrypoints.

### Docs

- [README](https://raw.githubusercontent.com/JordanGunn/asi/refs/heads/master/README.md): Overview, execution model, and quick start.
- [Manifesto](https://raw.githubusercontent.com/JordanGunn/asi/refs/heads/master/docs/manifesto/.INDEX.md): Narrative framing—*why* ASI exists.
- [Design Documentation](https://raw.githubusercontent.com/JordanGunn/asi/refs/heads/master/docs/design/.INDEX.md): Execution model, principles, and specifications.
- [Skills](https://raw.githubusercontent.com/JordanGunn/asi/refs/heads/master/docs/SKILLS.md): The skill design pipeline.
- [Glossary](https://raw.githubusercontent.com/JordanGunn/asi/refs/heads/master/docs/GLOSSARY.md): Key terms and definitions.