Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Phase-gated commit workflow for clean git history -- implement, review, test, commit per phase
.claude/skills/phase-gated-commits/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | — | — |
| case-21 | ✗→✓ | ▲ Improved | — | — |
| case-09 | ✗→✓ | ▲ Improved | — | — |
| case-14 | ✗→✓ | ▲ Improved | — | — |
| case-10 | ✗→✓ | ▲ Improved | — | — |
Break large features and refactoring into discrete phases. Each phase follows a strict implement -> review -> test -> commit cycle. The result is a clean, bisectable git history where every commit represents a working state.
Break the work into 2-5 phases. Each phase should:
| Phase Size | Guideline | |-----------|-----------| | Too small | Renaming a single variable is not a phase | | Right size | Add the data model + migration + basic tests | | Too large | Implement entire feature end-to-end in one phase |
Aim for phases that touch 2-8 files each. If a phase touches more than 10 files, consider splitting it further.
Phase N:
1. IMPLEMENT -- Write the code for this phase only
2. REVIEW -- Self-review the diff, check for issues
3. TEST -- Run tests, verify nothing is broken
4. COMMIT -- Create a single commit for this phase
5. PAUSE -- Wait for user confirmation before Phase N+1Write only the code that belongs to this phase. Do not reach ahead into the next phase. If you realize the current phase needs to be larger, stop and re-scope before continuing.
Review your own diff before committing:
If review finds issues, fix them within the same phase. Do not defer fixes to a later phase.
Run the project test suite. At minimum:
Create one commit per phase with a descriptive message.
Commit message format:
feat(phase N/M): <description of what this phase accomplishes>
<Optional body explaining WHY this phase exists as a separate unit>Examples:
feat(phase 1/3): add user preference data model and migration
Separate from the API layer so the schema can be reviewed independently.
feat(phase 2/3): implement preference API endpoints with validation
Builds on the data model from phase 1. Includes input validation
with zod schemas and error handling for all edge cases.
feat(phase 3/3): add preference UI components and integration tests
Connects the API to the frontend. Integration tests cover the
full create/read/update flow.After committing, pause and confirm with the user before starting the next phase. This gives the user a chance to:
When using this skill alongside plan-documentation:
<task>-phase-N-complete.md) are written after the commit| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-23 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-24 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +42 percentage points is the difference between those two pass rates over the 24 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.