Agent Skills: Housekeeping (AGENTS/CLAUDE Architecture)

Repository housekeeping workflows for AGENTS/CLAUDE architecture, progressive disclosure, and migration of legacy monolithic instruction files.

UncategorizedID: bout3fiddy/agents/housekeeping

Install this agent skill to your local

pnpm dlx add-skill https://github.com/bout3fiddy/agents/tree/HEAD/skills/housekeeping

Skill Files

Browse the full folder contents for housekeeping.

Download Skill

Loading file tree…

skills/housekeeping/SKILL.md

Skill Metadata

Name
housekeeping
Description
Repository housekeeping workflows for AGENTS/CLAUDE architecture, progressive disclosure, and migration of legacy monolithic instruction files.

Housekeeping (AGENTS/CLAUDE Architecture)

Use this skill when work is about organizing or modernizing AGENTS.md/CLAUDE.md context systems.

Operating rules

  • Keep root AGENTS.md concise: critical guardrails, task router, canonical commands, and links.
  • Prefer scoped instruction files in dense areas (apps/frontend, infra, apps/agent, etc.).
  • Move deep or volatile details into .agents/repo-context/*, not root AGENTS.md.
  • Curate continuously: deduplicate, resolve contradictions, and remove stale guidance.
  • Preserve behavior intent during migrations; change structure first, then tighten wording.

Workflow

  1. Assess current state (concise router vs mixed/legacy monolith).
  2. Select target architecture from references.
  3. Migrate content into scoped AGENTS.md files and deep docs.
  4. Rewrite root AGENTS.md as a short router.
  5. Add freshness metadata for volatile facts where feasible.
  6. Verify links/routes and remove conflicting duplicate rules.
  7. If work includes both AGENTS.md and CLAUDE.md, apply parity checks in the same pass.

Canonical ownership

  • instructions/global.md: trigger registration, precedence, escalation, and lifecycle policy.
  • skills/housekeeping/SKILL.md: workflow sequence and reference selection for housekeeping execution.
  • skills/housekeeping/references/*: deep guidance and operating patterns; AGENTS and CLAUDE are references.

Reference triggers (open when clearly relevant)

  • AGENTS/CLAUDE architecture, progressive disclosure, or context-memory organization -> skills/housekeeping/references/agents-architecture.md
  • Legacy monolithic AGENTS cleanup, contradiction pruning, or migration planning -> skills/housekeeping/references/migration-playbook.md

References

  • skills/housekeeping/references/index.md - References index for this skill
  • skills/housekeeping/references/agents-architecture.md - Target architecture, file tree, and root template
  • skills/housekeeping/references/migration-playbook.md - Legacy-to-modern migration workflow and checklist