General-purpose skills for AI agents.
This repository collects practical, reusable instructions that help agents give better responses with fewer tokens. Skills should improve common workflows, reduce repeated reasoning, and stay useful across projects and domains.
- Broadly useful, not niche
- Clear enough to follow consistently
- Compact enough to justify their context cost
- Focused on better outcomes, not longer prompts
Each skill lives in its own directory with a SKILL.md file and any supporting
assets it needs.
I recommend using the Releases section in the sidebar to download the specific
skill(s) you want as a zip file. Then, simply unzip it to user or repository scope.
See the official Codex skills documentation for supported locations, symlinks, and skill discovery behavior.
Note that individual skill releases are not independently versioned. A version bump on any skill bumps every skill. Just because a skill has a new version number, that doesn't mean it's been changed.
Keeps project facts, implementation decisions, validation findings, and known
limits in a version-controlled source of truth at memory/MEMORY.md.
Agents using this skill:
- Read the relevant memory entry before changing a subsystem.
- Update that entry after making the change.
- Link entries to the related project files and explain what those files do.
- Remove stale behavior instead of preserving history; memory describes the current state and is not a changelog.
Future work belongs under TODO / Future Projects. New entries can start from
the included assets/template.md and adapt
its sections to the subsystem being documented.
See intree-memory/SKILL.md for the complete rules.