curator

name	curator
description	Maintain Packy's repo-local skill library by creating, revising, consolidating, and pruning skills. Use when Packy work reveals a repeated workflow, a durable procedure, stale skill guidance, or a lesson that may belong in SKILLS.md or a skill folder instead of only AGENTS.md or MEMORY.md.

Curator

Overview

Use this skill to keep Packy's reusable agent procedures small, accurate, and discoverable. The goal is a compact skill library that helps future agents repeat real workflows without turning every one-off note into a skill.

Workflow

1. Decide whether the lesson is reusable.
   • Put project-wide operating rules in AGENTS.md.
   • Put durable observations, collaborator cues, and one-off findings in MEMORY.md.
   • Put repeated procedures with steps, validations, or assets in skills/<name>/SKILL.md.
2. Search SKILLS.md and skills/ before creating anything new.
3. Update an existing skill when the workflow is already covered.
4. Create a new skill only when the procedure has a clear trigger and future use.
5. Prune or consolidate stale skills when they overlap or no longer match Packy's workflow.
6. Update SKILLS.md whenever the available skill set changes.
7. Keep agents/openai.yaml aligned with the skill name, purpose, and default prompt.
8. Validate before committing:

python3 /home/pierce/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/<name>
git diff --check

Skill Standards

• Keep frontmatter to name and description.
• Put all trigger guidance in the description so agents can decide whether to load the skill.
• Keep the body procedural and concise.
• Add scripts/, references/, or assets/ only when they directly support the skill.
• Delete generated TODO text before committing.
• Do not create auxiliary README, changelog, or quick-reference files inside a skill.

Naming

Use lowercase hyphenated names under 64 characters. Prefer short names that describe the action or workflow, such as publish-site-change or package-install-audit.

Commit Criteria

Before landing skill changes, the relevant skill validates, SKILLS.md lists the current skill set, and AGENTS.md or MEMORY.md reflects any broader operating lesson.