Agent Docs

The agent-NATIVE documentation an AI agent reads and acts on — AGENTS.md, llms.txt, tool descriptions, machine-readable reference, and the context budget. Narrowed to artifacts with no human-docs analog; generic and dual-audience docs stay in docs-audit / docs-design. Provenance lives in skill.json; this file is runtime routing only.

Produces: a change-plan.md (DO), an audit-report.md plus a findings-ledger + workflow-state when tracked (REVIEW), or a design-doc.md / refactor-runbook.md / explanation.md (DESIGN).

Core principle

Write so an agent can find it, use a chunk of it alone, trigger on it correctly, and not drown in it. Curation raises average success, not worst-case reliability — mandatory invariants belong in gates (route to harden-repo-for-coding-agents), not ignorable prose.

Activation