📋 Full Skill Source — This is the complete, unedited SKILL.md file. Nothing is hidden or summarized.
Skill Mastery — Use + Create + Discover
The meta-skill: How to find, use, and create cm-* skills.
Part A: Using Skills
The Rule
Invoke relevant skills BEFORE any response or action. Even 1% chance = check the skill.
Decision Flow
User message received
→ Might any skill apply? (even 1%)
→ YES: Read the skill → Follow it
→ NO: Respond directlySkill Priority
- Process skills first (cm-planning, cm-debugging) — determine HOW to approach
- Implementation skills second (cm-tdd, cm-safe-deploy) — guide execution
Red Flags
| Thought | Reality |
|---|---|
| "This is just a simple question" | Questions are tasks. Check for skills. |
| "I need more context first" | Skill check BEFORE clarifying questions. |
| "The skill is overkill" | Simple things become complex. Use it. |
| "I remember this skill" | Skills evolve. Read current version. |
| "Let me just do this one thing first" | Check BEFORE doing anything. |
Skill Types
- Rigid (cm-tdd, cm-debugging): Follow exactly. Don't adapt away discipline.
- Flexible (cm-planning): Adapt principles to context.
Part B: Creating Skills
When to Create
- Pattern repeated across 3+ projects
- Complex process that needs documentation
- Team convention that should be enforced
Skill Structure
markdown
---
name: cm-[skill-name]
description: "One line: when to use this skill"
---
# [Skill Title]
## Overview
What and why. Core principle.
## When to Use
Specific triggers.
## The Process
Step-by-step instructions.
## Red Flags
What NOT to do.
## Integration
How it connects to other cm-* skills.Rules
✅ DO:
- Keep under 400 lines (token optimization)
- Use tables and code blocks for density
- Include Red Flags section
- Reference other cm-* skills by name
- Test skill with real scenarios before deploying
❌ DON'T:
- Write prose when a table works
- Duplicate content from other skills (reference instead)
- Create skills for one-time tasks
- Exceed 600 lines without good reasonToken Optimization
| Technique | Saves |
|---|---|
| Tables over prose | ~40% |
| Code blocks over explanation | ~30% |
| Reference other skills vs duplicate | ~50% |
| Remove obvious examples | ~20% |
Part C: Discovering Skills
Adaptive Skills Discovery
When you encounter something you don't have a skill for:
1. DETECT → "I need to do X but no matching skill"
2. SEARCH → npx skills find "{keyword}"
3. REVIEW → Read the SKILL.md — safe? relevant?
4. ASK → "Found skill '{name}'. Install?"
5. INSTALL → npx skills add {source} --skill {name} -a antigravity
6. USE → Apply the skill
7. LOG → Record in .cm-skills-log.jsonSafety Rules
- Always show user what you found before installing
- Prefer known repos (vercel-labs/agent-skills)
- Project-level by default, global only if agreed
- Check
npx skills listto avoid duplicates - Never override existing cm-* skills with external ones
The Cody Master Kit (13 skills)
| # | Skill | Purpose |
|---|---|---|
| 1 | cm-project-bootstrap | Start new projects |
| 2 | cm-planning | Brainstorm + write plans |
| 3 | cm-execution | Execute plans (batch/subagent/parallel) |
| 4 | cm-tdd | Test-driven development |
| 5 | cm-quality-gate | Test + verify + frontend safety |
| 6 | cm-code-review | Request + receive + finish branch |
| 7 | cm-safe-deploy | 6-gate deploy pipeline |
| 8 | cm-safe-i18n | i18n batching + safety |
| 9 | cm-debugging | Systematic root-cause debugging |
| 10 | cm-terminal | Terminal command monitoring |
| 11 | cm-git-worktrees | Isolated workspaces |
| 12 | cm-skill-mastery | This skill (meta) |
| 13 | cm-identity-guard | Git/deploy identity safety |
The Bottom Line
Skills are discipline, not overhead. Use them. Create them. Never skip them.