| name | aigis |
| description | AI governance for coding agents. Three-layer skill system — core / area / workflow. Use the aigis CLI to load skills on demand. Mapped to NIST AI RMF, OWASP Top 10 for LLMs, and ISO/IEC 42001. |
Aigis skill system
Aigis ships skills in three layers. Agents load them on demand via the CLI, not by reading this file.
Layers
- Core skill —
content/skills/core.md. Always loaded into the agent's context by aigis init <ide>. Orients the agent to the workflow + the four CLI commands. ~50 lines.
- Area skills —
content/skills/areas/<id>.md, one per governance area (input-validation, pii-handling, etc.). 15 today. Loaded by aigis get <id> when classify points the agent at that area.
- Workflow skills —
content/skills/workflows/<type>.md. Loaded by aigis workflow <type> after classify identifies the project shape. Step 5 ships fastapi-llm; future workflows: express-llm, nextjs-llm, agentic, rag-service.
How an agent uses Aigis
After aigis init writes the core skill into the agent's rules file, the agent runs:
aigis classify — get risk tier + recommended areas + workflow suggestion
aigis workflow <type> — get the canonical file layout + wiring contracts
aigis get <area> and aigis verify <area> --auto . per area, in order
See content/skills/core.md for the full prose version that the agent sees.
Setup per IDE
Run once in the project root:
aigis init cursor — writes .cursorrules
aigis init claude-code — writes ~/.claude/skills/aigis/SKILL.md
aigis init windsurf — writes .windsurfrules
aigis init copilot — writes .github/copilot-instructions.md
All four destinations receive the same content: content/skills/core.md. Area and workflow skills load on demand; they are not pre-installed into the rules file.