Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review C++ changes for string parameter and call-site efficiency conventions (`std::string_view`, `std::string&&`, `const std::string&`, `const char*`, and TransparentStringMap lookup). Use when reviewing C++ code, refactoring addon interfaces, or invoking /cpp-string-api-review.
.claude/skills/tetherto-cpp-string-api-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | -14% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | -22% | 0% |
Review C++ code for string API convention compliance and produce actionable fixes.
Use when:
/cpp-string-api-review.Do NOT use for:
docs/conventions/string-api-conventions.mdpackages/inference-addon-cpp/src/inference-addon-cpp/TransparentStringMap.hpp*.h, *.hpp, *.cc, *.cpp, *.cxx) in scope.std::string_view),std::string&&),const std::string&),const char*).const std::string& where std::string_view is sufficient,const std::string& + copy instead of std::string&& + move,std::string{view}) on hot paths,view.data() usage where null termination is not guaranteed,std::string_view but must call a const std::string& API, surface the trade-off and propose the correct boundary.Use this structure:
markdown## C++ String API Review ### Findings - [High|Medium|Low] <short title> — <file/symbol> - Why it matters: <cost/correctness impact> - Suggested change: <concise fix> ### Recommended edits 1. <edit 1> 2. <edit 2> ### Notes - <boundary trade-offs, if any>
std::string_view through call chains.const std::string&, do not hide that cost behind wrapper layers.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 23,618 | 6,073 | -74% | 1 | 1 | 0% | 4,434 | 1,663 | -62% | 0 | 0 | — |
case-02 | fail→fail | 16,316 | 21,550 | +32% | 1 | 1 | 0% | 2,830 | 4,577 | +62% | 0 | 0 | — |
case-03 | fail→fail | 20,433 | 5,271 | -74% | 1 | 1 | 0% | 3,636 | 1,498 | -59% | 0 | 0 | — |
case-04 | fail→fail | 13,224 | 8,904 | -33% | 1 | 1 | 0% | 2,334 | 2,255 | -3% | 0 | 0 | — |
case-05 | pass→pass | 9,705 | 4,427 | -54% | 1 | 1 | 0% | 1,708 | 1,466 | -14% | 0 | 0 | — |
case-06 | pass→pass | 10,704 | 4,513 | -58% | 1 | 1 | 0% | 1,759 | 1,369 | -22% | 0 | 0 | — |
case-07 | pass→pass | 12,407 | 9,272 | -25% | 1 | 1 | 0% | 2,266 | 2,359 | +4% | 0 | 0 | — |
case-08 | pass→pass | 11,232 | 8,230 | -27% | 1 | 1 | 0% | 2,029 | 2,080 | +3% | 0 | 0 | — |
case-09 | pass→pass | 10,762 | 5,425 | -50% | 1 | 1 | 0% | 1,713 | 1,473 | -14% | 0 | 0 | — |
case-10 | pass→pass | 10,421 | 7,727 | -26% | 1 | 1 | 0% | 1,806 | 2,041 | +13% | 0 | 0 | — |
case-11 | fail→pass | 10,292 | 3,957 | -62% | 1 | 1 | 0% | 1,742 | 1,312 | -25% | 0 | 0 | — |
case-12 | pass→pass | 8,145 | 4,852 | -40% | 1 | 1 | 0% | 1,326 | 1,447 | +9% | 0 | 0 | — |
case-13 | pass→pass | 7,556 | 3,838 | -49% | 1 | 1 | 0% | 1,196 | 1,220 | +2% | 0 | 0 | — |
case-14 | pass→pass | 2,852 | 9,800 | +244% | 1 | 1 | 0% | 426 | 1,792 | +321% | 0 | 0 | — |
case-15 | fail→pass | 14,834 | 11,480 | -23% | 1 | 1 | 0% | 2,482 | 2,610 | +5% | 0 | 0 | — |
case-16 | pass→pass | 9,245 | 7,777 | -16% | 1 | 1 | 0% | 1,683 | 1,988 | +18% | 0 | 0 | — |
case-17 | pass→pass | 13,880 | 9,645 | -31% | 1 | 1 | 0% | 2,326 | 2,186 | -6% | 0 | 0 | — |
case-18 | pass→pass | 7,585 | 5,200 | -31% | 1 | 1 | 0% | 1,292 | 1,470 | +14% | 0 | 0 | — |
case-19 | fail→pass | 11,244 | 8,173 | -27% | 1 | 1 | 0% | 1,840 | 1,937 | +5% | 0 | 0 | — |
case-20 | pass→pass | 17,448 | 13,842 | -21% | 1 | 1 | 0% | 3,705 | 3,309 | -11% | 0 | 0 | — |
case-21 | pass→pass | 13,822 | 9,347 | -32% | 1 | 1 | 0% | 2,396 | 2,231 | -7% | 0 | 0 | — |
case-22 | pass→pass | 14,918 | 8,236 | -45% | 1 | 1 | 0% | 2,592 | 2,146 | -17% | 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. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 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.