Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Verifies workspace state and staged changes as a read-only preflight. Use before commits or PRs to confirm staged set is clean and correct.
.claude/skills/athola-git-workspace-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 303% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 136% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -8% | 0% |
sanctum:commit-messages)sanctum:pr-prep)Run git status after review to verify workspace state matches expectations.
Run pytest plugins/sanctum/tests/test_git_workspace_review.py to validate review workflow.
Use this skill before workflows that depend on repository state, such as commit message generation, PR preparation, or release notes. Run it once per session or whenever staged changes are modified.
git-review:repo-confirmedgit-review:status-overviewgit-review:code-quality-checkgit-review:diff-statgit-review:diff-detailsMark each item as complete as you finish the corresponding step.
repo-confirmed)Run pwd to confirm you are in the correct repository directory. Execute git status -sb to view the current branch and short status, then capture the branch name and upstream information.
status-overview)Analyze the git status -sb output for staged and unstaged changes. Stage or unstage files so that subsequent workflows operate on the intended diff.
code-quality-check)Run make format && make lint to validate code quality before committing. Fix any errors immediately. Do not bypass pre-commit hooks with --no-verify. This check identifies issues early and avoids late-stage pipeline failures.
diff-stat)Run git diff --cached --stat for staged changes (or git diff --stat for unstaged work). Note the number of files modified and identify hotspots with large insertion or deletion counts.
When sem is available (see leyline:sem-integration), also run sem diff --format plain --staged to display an entity-level summary alongside the stat output. This shows which functions, classes, and methods changed rather than just line counts.
diff-details)Run git diff --cached to examine the actual changes. For unstaged work, use git diff. Identify key themes, such as Makefile adjustments or new skill additions, to provide context for downstream summaries.
Complete all progress tracking items. You should have a clear understanding of modified files and areas, and the correct work should be staged. Subsequent workflows can then rely on this context without re-executing git commands.
If pre-commit hooks block a commit, resolve the reported issues instead of using --no-verify. Run make format to fix styling errors automatically and use make lint to isolate logical failures. If merge conflicts occur, use git merge --abort to return to a clean state before retrying.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,874 | 6,463 | -6% | 1 | 1 | 0% | 1,134 | 962 | -15% | 0 | 0 | — |
case-02 | fail→fail | 3,277 | 3,851 | +18% | 1 | 1 | 0% | 475 | 1,229 | +159% | 0 | 0 | — |
case-03 | fail→fail | 6,283 | 7,175 | +14% | 1 | 1 | 0% | 289 | 1,250 | +333% | 0 | 0 | — |
case-04 | fail→fail | 4,782 | 6,652 | +39% | 1 | 1 | 0% | 772 | 1,591 | +106% | 0 | 0 | — |
case-05 | fail→pass | 2,418 | 7,394 | +206% | 1 | 1 | 0% | 340 | 1,371 | +303% | 0 | 0 | — |
case-06 | fail→pass | 7,247 | 14,689 | +103% | 1 | 1 | 0% | 1,041 | 2,456 | +136% | 0 | 0 | — |
case-07 | fail→pass | 4,349 | 2,455 | -44% | 1 | 1 | 0% | 660 | 1,056 | +60% | 0 | 0 | — |
case-08 | fail→pass | 10,985 | 3,934 | -64% | 1 | 1 | 0% | 1,744 | 1,284 | -26% | 0 | 0 | — |
case-09 | fail→pass | 9,102 | 4,306 | -53% | 1 | 1 | 0% | 1,471 | 1,353 | -8% | 0 | 0 | — |
case-10 | pass→pass | 5,809 | 2,595 | -55% | 1 | 1 | 0% | 967 | 1,085 | +12% | 0 | 0 | — |
case-11 | pass→pass | 3,841 | 2,548 | -34% | 1 | 1 | 0% | 583 | 1,104 | +89% | 0 | 0 | — |
case-12 | fail→pass | 13,584 | 2,664 | -80% | 1 | 1 | 0% | 2,275 | 1,142 | -50% | 0 | 0 | — |
case-13 | pass→pass | 2,785 | 1,972 | -29% | 1 | 1 | 0% | 419 | 1,033 | +147% | 0 | 0 | — |
case-14 | pass→pass | 9,267 | 2,518 | -73% | 1 | 1 | 0% | 1,433 | 1,063 | -26% | 0 | 0 | — |
case-15 | pass→pass | 3,184 | 1,549 | -51% | 1 | 1 | 0% | 429 | 911 | +112% | 0 | 0 | — |
case-16 | fail→pass | 9,008 | 2,734 | -70% | 1 | 1 | 0% | 1,518 | 1,164 | -23% | 0 | 0 | — |
case-17 | pass→pass | 5,501 | 1,882 | -66% | 1 | 1 | 0% | 928 | 1,029 | +11% | 0 | 0 | — |
case-18 | pass→pass | 8,672 | 3,452 | -60% | 1 | 1 | 0% | 1,372 | 1,200 | -13% | 0 | 0 | — |
case-19 | pass→pass | 11,969 | 3,248 | -73% | 1 | 1 | 0% | 1,808 | 1,277 | -29% | 0 | 0 | — |
case-20 | fail→fail | 10,741 | 1,901 | -82% | 1 | 1 | 0% | 1,694 | 922 | -46% | 0 | 0 | — |
case-21 | fail→pass | 13,425 | 5,698 | -58% | 1 | 1 | 0% | 2,118 | 1,562 | -26% | 0 | 0 | — |
case-22 | pass→pass | 7,820 | 1,898 | -76% | 1 | 1 | 0% | 1,199 | 919 | -23% | 0 | 0 | — |
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. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.