Standards for self-documenting code and minimal comments. Use when auditing inline documentation or deciding when comments add value vs clutter.
Install
npx skillscat add rileyhilliard/claude-essentials/documenting-code-comments Install via the SkillsCat registry.
SKILL.md
Code Comments
Core principle: The best comment is the one you didn't need to write.
Hierarchy
- Make code self-documenting (naming, structure)
- Use type systems for contracts
- Add comments only for WHY, never WHAT
When NOT to Comment
| Avoid | Why |
|---|---|
// Get the user's name |
Restates code |
@param {string} email |
Types already document |
| Stale comments | Misleading > missing |
When TO Comment
WHY, Not WHAT
// Use exponential backoff - service rate-limits after 3 rapid failures
const backoffMs = Math.pow(2, attempts) * 1000;Gotchas and Edge Cases
// IMPORTANT: Assumes UTC - local timezone causes date drift
const dayStart = new Date(date.setHours(0, 0, 0, 0));External Context
// Workaround for Safari flexbox bug (JIRA-1234)
// Per RFC 7231 §6.5.4, return 404 for missing resourcesPerformance Decisions
// Map for O(1) lookup - benchmarked 3x faster than array.find() at n>100
const userMap = new Map(users.map(u => [u.id, u]));Refactor Before Commenting
| Instead of comment | Refactor to |
|---|---|
// Get active users |
const activeUsers = users.filter(u => u.isActive) |
// 86400000 ms = 1 day |
const ONE_DAY_MS = 24 * 60 * 60 * 1000 |
// Handle error case |
Extract to handleAuthError(err) |
TODO Format
// ✅ TODO(JIRA-567): Replace with batch API when available Q1 2025
// ❌ TODO: fix this laterAudit Checklist
- Necessity - Can code be refactored to eliminate comment?
- Accuracy - Does comment match current behavior?
- Value - Does it explain WHY, not WHAT?
- Actionability - TODOs have ticket references?