Skip to content

AGENTS.md

Latest commit

 

History

History
22 lines (18 loc) · 2.33 KB

AGENTS.md

File metadata and controls

22 lines (18 loc) · 2.33 KB

This repository contains a small working set of agent skills and support assets that follow the agentskills.io style: lean SKILL.md files, progressive disclosure through references/ and assets/, and deterministic helper scripts where guessing would be brittle.

Repository layout:

  • skills/* primary repository content: each subfolder is a standalone skill with its own SKILL.md, scripts/, references/, and assets/.
  • .github/skills/ is the local supporting skills used to create, update, validate, and review skills in this repository.
  • assets/ is for internal supporting materials only. Do not treat it as user-facing repository content and do not consider it for README updates.
  • artifacts/* might contain disposable demos for testing skills. Never use this folder for any persistent asset, because it is ignored by Git and may be deleted without warning.
  • .github/prompts/ contains maintenance prompts for validating skills, remediating skill issues, updating the Prompt API skill, and recreating the demo artifact.

When working in this repo:

  1. Treat skills as agent-facing control files, not human documentation bundles.
  2. In README sections named Included Skills, list only skills that exist under /skills/.
  3. Keep each skill directory flat under scripts/, references/, and assets/.
  4. Match the skill folder name and the YAML name field exactly.
  5. Keep SKILL.md lean and procedural; move bulky rules, examples, and templates into references/ or assets/.
  6. Use relative paths with forward slashes in skill files.
  7. Do not add per-skill human-oriented files such as README.md or CHANGELOG.md.
  8. When creating or revising a skill, use .agents/skills/skill-creator/scripts/validate-metadata.py and .agents/skills/skill-creator/references/checklist.md as the repository's validation path.
  9. For saved maintenance prompts that gather external information, request sources in this priority order unless the prompt explicitly says otherwise: supplied prompt text, supplied attached documents, then the built-in URLs already defined in the prompt. Higher-priority sources override lower-priority sources.
  10. Never consider contents of artifacts/ as persistent or canonical. Never add description of it to the README or reference it from any skill. Always treat it as a scratch space that can be deleted without warning.