agent-instructions
Use when creating, editing, tightening, or reviewing files that instruct agents — CLAUDE.md, AGENTS.md, SKILL.md, agent definition files, or system-prompt fragments — including reviewing a diff to such a file for instruction quality, and before writing the prompt that dispatches a sub-agent or delegated task. Provides the house style for unambiguous, auditable instructions, the document mechanics, and guidance for one-shot sub-agent prompts. Do not use as the entry point for reviewing a session's own changes before a commit or PR, which a review workflow owns and applies this skill inside; for technical artifacts such as READMEs or documentation (technical-writing); for code comments, commit messages, or chat responses; for maintaining a terminology section's terms, aliases, or relationships; for conceptual questions about instruction files; or for implementation requests the session codes directly, even when the program is an agent; delegating implementation to a sub-agent falls under the dispatch trigger.
- License
- MIT
Pinned to revision ae7c2bfbb448, so it is the text this page describes rather than whatever the author pushed since.
Files
- skills/agent-instructions/SKILL.md
- skills/agent-instructions/agents/openai.yaml
- skills/agent-instructions/evals/triggers.yaml
- skills/agent-instructions/references/agents-md.md
- skills/agent-instructions/references/document-mechanics.md
- skills/agent-instructions/references/review-checklist.md
- skills/agent-instructions/references/skill-mechanics.md
- skills/agent-instructions/references/subagent-prompts.md
Every link opens the file at its source, pinned to the revision this page describes.