Imported from powerpuff-kitty/agentic-harness (AGENTS.md). Install upstream with npx skills add powerpuff-kitty/agentic-harness. Copyright stays with the author.

Agent instructions

This repository is a self-hosting Agentic Harness project and the canonical catalog source.

1. Read .agentic/README.md for the project-context map.
2. Read .agentic/manifest.yaml for paths, modules, permissions, and source ownership.
3. Load only the relevant canonical truth, ADRs, supporting docs, packs, policies, and skills.
4. Apply precedence: policies; current project truth; accepted ADRs; packs; skills; one-off prompts.
5. catalog/ owns reusable variants and modules. Changes to the target-project contract must update schemas, validation, ADRs/current truth, agent procedures, CLI source pins, and migration tests as applicable.
6. Do not recreate legacy root-level project truth, docs/decisions, hidden overlay-only public variants, or duplicated vendor instructions.
7. Preserve project-authored truth, keep changes scoped and reversible, and never expose secrets or perform destructive/release actions without the approvals in the manifest.
8. Run catalog validation and relevant downstream tests before declaring a cross-repository change complete.
9. State unresolved compatibility, migration, security, and validation gaps explicitly.
10. Follow the public-surface policy for files and collaboration metadata; never identify internal consumers in public content or fixtures.
11. For TypeSafe/Jev integration, semantic decision features, or AI feature planning, use the installed TypeSafe skill and its live-documentation workflow. Keep deterministic rules and authorization in project-owned code.

Vendor adapters must remain thin and point back to this file and .agentic/; they may not redefine canonical architecture.