Skill Author

Inputs

• •Skill intent, example triggers, and any needed resources (scripts, references, assets).
• •Optional: target agent(s) and desired name.

Workflow

1. •Define skill name and triggers
   • •Use lowercase letters, digits, hyphens; keep under 64 chars.
   • •Ensure description captures when to use the skill (primary trigger signal).
2. •Create shared skill folder
   • •Path: master/skills/<skill-name>/SKILL.md
   • •Frontmatter: only name and description.
3. •Write SKILL.md body
   • •Imperative voice.
   • •Keep concise; avoid duplicate docs.
   • •Include guardrails and failure cases.
4. •Add resources if needed
   • •scripts/, references/, assets/ under the same folder.
5. •Wire into agents
   • •Codex: symlink codex/.codex/skills/<skill-name> -> master/skills/<skill-name>
   • •Claude: symlink claude/.claude/skills/<skill-name> -> master/skills/<skill-name>
6. •Optionally add Gemini wrapper
   • •.gemini/commands/<skill-name>.toml with @{master/skills/<skill-name>/SKILL.md}
7. •Sanity check
   • •Confirm symlinks and that SKILL.md stays under ~500 lines.

Guardrails

• •Do not add extra docs (README, changelog).
• •Avoid long examples; move detail into references/.
• •Ask before overwriting an existing skill.