Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audits Makefiles for build correctness, portability, and recipe duplication. Use when reviewing a Makefile or before committing Makefile changes.
.claude/skills/athola-makefile-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 194% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 31% | 0% |
makefile-review:context-mapped))makefile-review:dependency-graph))makefile-review:dedup-candidates))makefile-review:tooling-alignment))makefile-review:evidence-logged))Run pytest plugins/pensive/tests/skills/test_makefile_review.py to verify review logic.
Audit Makefiles for best practices, deduplication, and portability.
bash/makefile-review
makefile-review:context-mappedmakefile-review:dependency-graphmakefile-review:dedup-candidatesmakefile-review:tooling-alignmentmakefile-review:evidence-loggedmakefile-review:findings-verifiedmakefile-review:context-mapped)Confirm baseline:
bashpwd && git status -sb && git diff --stat
Verification: Run git status to confirm working tree state.
Find Make-related files:
bashrg -n "^include" -g'Makefile*' rg --files -g '*.mk'
Document changed targets, project goals, and tooling requirements.
makefile-review:dependency-graph)@include modules/dependency-graph.md
makefile-review:dedup-candidates)@include modules/deduplication-patterns.md
makefile-review:tooling-alignment)@include modules/portability-checks.md
makefile-review:evidence-logged)Use imbue:proof-of-work to record command outputs with file:line references.
Summarize findings:
Load additional context as needed:
Best Practices & Examples: @include modules/best-practices.md
Plugin Dogfood Checks: @include modules/plugin-dogfood-checks.md - Makefile completeness analysis, target generation, and dogfooding validation.
markdown## Summary Makefile review findings ## Context - Files reviewed: [list] - Targets changed: [list] ## Dependency Analysis [graph and issues] ## Duplication Candidates ### [D1] Repeated command - Locations: [list] - Anchor: `verbatim source text at file:line` - Recommendation: [pattern rule] ## Portability Issues [cross-platform concerns] ## Missing Targets - [ ] help - [ ] format - [ ] lint ## Recommendation Approve / Approve with actions / Block
makefile-review:findings-verified)Every finding must cite a real location and a verbatim anchor. Write findings to .review/findings.json and confirm each citation resolves:
bashpython plugins/imbue/scripts/citation_verifier.py \ --findings .review/findings.json --repo-root .
Drop or label UNVERIFIED any finding the verifier fails (exit 1); only verified findings enter the report. See Skill(imbue:review-core) Step 5 and Skill(imbue:structured-output) for the schema.
Location + verbatim Anchor confirmedby citation_verifier.py (exit 0), or unverified findings were dropped or labeled UNVERIFIED
No Makefile found Ensure Makefile or *.mk files exist in the project root or specify paths explicitly.
Include directives not resolved Run rg -n "^include" -g'Makefile*' to trace include chains manually.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 10,721 | 6,458 | -40% | 1 | 1 | 0% | 1,605 | 1,506 | -6% | 0 | 0 | — |
case-16 | fail→pass | 9,682 | 1,822 | -81% | 1 | 1 | 0% | 1,359 | 1,460 | +7% | 0 | 0 | — |
case-22 | fail→pass | 9,616 | 1,724 | -82% | 1 | 1 | 0% | 1,498 | 1,405 | -6% | 0 | 0 | — |
case-02 | fail→fail | 20,605 | 5,649 | -73% | 1 | 1 | 0% | 3,535 | 1,562 | -56% | 0 | 0 | — |
case-03 | fail→fail | 29,949 | 5,977 | -80% | 1 | 1 | 0% | 5,109 | 1,475 | -71% | 0 | 0 | — |
case-04 | pass→pass | 12,575 | 16,084 | +28% | 1 | 1 | 0% | 2,342 | 4,284 | +83% | 0 | 0 | — |
case-05 | pass→pass | 10,083 | 6,072 | -40% | 1 | 1 | 0% | 1,520 | 2,013 | +32% | 0 | 0 | — |
case-06 | pass→pass | 10,385 | 20,247 | +95% | 1 | 1 | 0% | 1,771 | 4,590 | +159% | 0 | 0 | — |
case-07 | fail→pass | 7,648 | 2,349 | -69% | 1 | 1 | 0% | 499 | 1,469 | +194% | 0 | 0 | — |
case-08 | fail→fail | 17,256 | 4,763 | -72% | 1 | 1 | 0% | 2,607 | 1,886 | -28% | 0 | 0 | — |
case-09 | fail→pass | 14,450 | 2,609 | -82% | 1 | 1 | 0% | 1,221 | 1,575 | +29% | 0 | 0 | — |
case-10 | fail→pass | 9,960 | 5,146 | -48% | 1 | 1 | 0% | 1,583 | 2,067 | +31% | 0 | 0 | — |
case-11 | fail→pass | 15,018 | 2,594 | -83% | 1 | 1 | 0% | 2,264 | 1,651 | -27% | 0 | 0 | — |
case-12 | pass→pass | 9,118 | 3,867 | -58% | 1 | 1 | 0% | 1,400 | 1,796 | +28% | 0 | 0 | — |
case-13 | fail→pass | 16,338 | 9,243 | -43% | 1 | 1 | 0% | 2,568 | 1,898 | -26% | 0 | 0 | — |
case-14 | fail→pass | 12,345 | 1,721 | -86% | 1 | 1 | 0% | 1,870 | 1,412 | -24% | 0 | 0 | — |
case-15 | fail→pass | 11,114 | 2,298 | -79% | 1 | 1 | 0% | 1,707 | 1,568 | -8% | 0 | 0 | — |
case-17 | fail→pass | 6,595 | 2,218 | -66% | 1 | 1 | 0% | 967 | 1,478 | +53% | 0 | 0 | — |
case-18 | fail→pass | 2,773 | 2,438 | -12% | 1 | 1 | 0% | 344 | 1,508 | +338% | 0 | 0 | — |
case-19 | fail→pass | 7,052 | 1,672 | -76% | 1 | 1 | 0% | 1,084 | 1,449 | +34% | 0 | 0 | — |
case-20 | fail→pass | 5,742 | 2,156 | -62% | 1 | 1 | 0% | 789 | 1,533 | +94% | 0 | 0 | — |
case-21 | pass→pass | 10,752 | 2,648 | -75% | 1 | 1 | 0% | 1,644 | 1,659 | +1% | 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 19 counted toward the lift figure. The other 3 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 +59 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.