Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert C++ code reviewer specializing in memory safety, modern C++ idioms, concurrency, and performance. Use for all C++ code changes. MUST BE USED for C++ projects.
.claude/skills/kunanonj-agent-cpp-reviewer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 85% | 0% |
You are a senior C++ code reviewer ensuring high standards of modern C++ and best practices.
When invoked:
git diff -- '*.cpp' '*.hpp' '*.cc' '*.hh' '*.cxx' '*.h' to see recent C++ file changesclang-tidy and cppcheck if availablestd::unique_ptr or std::shared_ptrstrcpy, sprintf without boundssystem() or popen()printf format stringreinterpret_cast without justificationlock()/unlock() instead of std::lock_guardstd::thread without join() or detach()malloc, C arrays, typedef instead of usingconst&std::move for sink parametersstd::ostringstream or reserve()reserve(): Known-size vector without pre-allocationconst correctness: Missing const on methods, parameters, referencesauto overuse/underuse: Balance readability with type deductionusing namespace std; in headersbashclang-tidy --checks='*,-llvmlibc-*' src/*.cpp -- -std=c++17 cppcheck --enable=all --suppress=missingIncludeSystem src/ cmake --build build 2>&1 | head -50
For detailed C++ coding standards and anti-patterns, see skill: cpp-coding-standards.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | pass→pass | 7,468 | 5,276 | -29% | 1 | 1 | 0% | 1,423 | 1,920 | +35% | 0 | 0 | — |
case-09 | pass→pass | 12,921 | 10,101 | -22% | 1 | 1 | 0% | 2,452 | 2,552 | +4% | 0 | 0 | — |
case-01 | fail→fail | 15,604 | 12,479 | -20% | 1 | 1 | 0% | 1,691 | 2,421 | +43% | 0 | 0 | — |
case-02 | fail→fail | 2,168 | 6,600 | +204% | 1 | 1 | 0% | 354 | 1,988 | +462% | 0 | 0 | — |
case-03 | fail→fail | 11,728 | 12,642 | +8% | 1 | 1 | 0% | 1,130 | 2,353 | +108% | 0 | 0 | — |
case-23 | pass→pass | 11,968 | 12,711 | +6% | 1 | 1 | 0% | 2,302 | 3,542 | +54% | 0 | 0 | — |
case-04 | fail→pass | 12,521 | 7,559 | -40% | 1 | 1 | 0% | 1,995 | 2,109 | +6% | 0 | 0 | — |
case-05 | pass→pass | 8,520 | 9,404 | +10% | 1 | 1 | 0% | 1,546 | 2,127 | +38% | 0 | 0 | — |
case-06 | pass→pass | 9,699 | 6,359 | -34% | 1 | 1 | 0% | 1,207 | 2,108 | +75% | 0 | 0 | — |
case-07 | pass→pass | 9,249 | 4,877 | -47% | 1 | 1 | 0% | 1,376 | 1,813 | +32% | 0 | 0 | — |
case-10 | fail→pass | 11,961 | 3,585 | -70% | 1 | 1 | 0% | 2,027 | 1,538 | -24% | 0 | 0 | — |
case-11 | pass→pass | 12,063 | 3,563 | -70% | 1 | 1 | 0% | 1,920 | 1,609 | -16% | 0 | 0 | — |
case-12 | fail→pass | 8,564 | 5,451 | -36% | 1 | 1 | 0% | 1,635 | 1,871 | +14% | 0 | 0 | — |
case-13 | pass→pass | 6,089 | 4,030 | -34% | 1 | 1 | 0% | 1,085 | 1,622 | +49% | 0 | 0 | — |
case-14 | fail→pass | 12,796 | 4,406 | -66% | 1 | 1 | 0% | 2,350 | 1,782 | -24% | 0 | 0 | — |
case-15 | fail→fail | 11,521 | 9,515 | -17% | 1 | 1 | 0% | 2,117 | 2,393 | +13% | 0 | 0 | — |
case-16 | pass→pass | 5,280 | 3,074 | -42% | 1 | 1 | 0% | 917 | 1,355 | +48% | 0 | 0 | — |
case-17 | fail→pass | 4,671 | 3,346 | -28% | 1 | 1 | 0% | 799 | 1,476 | +85% | 0 | 0 | — |
case-18 | pass→pass | 5,870 | 3,785 | -36% | 1 | 1 | 0% | 1,106 | 1,491 | +35% | 0 | 0 | — |
case-19 | fail→fail | 10,127 | 4,849 | -52% | 1 | 1 | 0% | 1,754 | 1,701 | -3% | 0 | 0 | — |
case-20 | pass→pass | 3,967 | 3,883 | -2% | 1 | 1 | 0% | 731 | 1,470 | +101% | 0 | 0 | — |
case-21 | pass→pass | 5,453 | 7,636 | +40% | 1 | 1 | 0% | 1,013 | 2,371 | +134% | 0 | 0 | — |
case-22 | pass→pass | 12,830 | 9,417 | -27% | 1 | 1 | 0% | 2,575 | 2,372 | -8% | 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. 23 cases were attempted. The headline lift of +22 percentage points is the difference between those two pass rates over the 23 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.