Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Execute macOS code signing with Developer ID and hardened runtime requirements
.claude/skills/a5c-ai-macos-codesign-workflow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 77% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 69% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 8% | 0% |
Execute macOS code signing with Developer ID certificates and hardened runtime. This skill handles the complete code signing process for macOS applications.
json{ "type": "object", "properties": { "appPath": { "type": "string" }, "identity": { "type": "string" }, "entitlements": { "type": "string" }, "hardenedRuntime": { "type": "boolean", "default": true } }, "required": ["appPath", "identity"] }
bash# Sign app bundle with hardened runtime codesign --force --options runtime --timestamp \ --entitlements MyApp.entitlements \ --sign "Developer ID Application: Company Name (TEAMID)" \ MyApp.app # Sign nested components first find MyApp.app -name "*.dylib" -o -name "*.framework" | \ xargs -I {} codesign --force --options runtime --timestamp \ --sign "Developer ID Application: Company Name (TEAMID)" {} # Verify signature codesign --verify --deep --strict --verbose=2 MyApp.app spctl --assess --type execute --verbose MyApp.app
macos-notarization-workflowmacos-entitlements-generator| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 9,866 | 7,763 | -21% | 1 | 1 | 0% | 1,725 | 1,864 | +8% | 0 | 0 | — |
case-02 | pass→pass | 5,771 | 2,681 | -54% | 1 | 1 | 0% | 883 | 798 | -10% | 0 | 0 | — |
case-03 | pass→pass | 6,185 | 4,005 | -35% | 1 | 1 | 0% | 955 | 1,026 | +7% | 0 | 0 | — |
case-04 | pass→pass | 5,329 | 6,016 | +13% | 1 | 1 | 0% | 1,020 | 1,517 | +49% | 0 | 0 | — |
case-05 | pass→pass | 5,818 | 2,631 | -55% | 1 | 1 | 0% | 1,027 | 791 | -23% | 0 | 0 | — |
case-06 | pass→pass | 3,564 | 2,049 | -43% | 1 | 1 | 0% | 535 | 735 | +37% | 0 | 0 | — |
case-07 | pass→pass | 10,400 | 1,872 | -82% | 1 | 1 | 0% | 1,886 | 695 | -63% | 0 | 0 | — |
case-08 | fail→pass | 9,299 | 3,519 | -62% | 1 | 1 | 0% | 1,716 | 1,045 | -39% | 0 | 0 | — |
case-09 | pass→pass | 10,233 | 1,466 | -86% | 1 | 1 | 0% | 1,833 | 575 | -69% | 0 | 0 | — |
case-10 | fail→pass | 2,447 | 1,958 | -20% | 1 | 1 | 0% | 405 | 717 | +77% | 0 | 0 | — |
case-11 | pass→pass | 2,037 | 1,819 | -11% | 1 | 1 | 0% | 331 | 674 | +104% | 0 | 0 | — |
case-12 | pass→pass | 4,384 | 2,104 | -52% | 1 | 1 | 0% | 793 | 707 | -11% | 0 | 0 | — |
case-13 | pass→pass | 2,907 | 2,720 | -6% | 1 | 1 | 0% | 469 | 779 | +66% | 0 | 0 | — |
case-14 | pass→pass | 10,974 | 4,751 | -57% | 1 | 1 | 0% | 1,890 | 1,160 | -39% | 0 | 0 | — |
case-15 | pass→pass | 11,222 | 4,097 | -63% | 1 | 1 | 0% | 1,958 | 1,095 | -44% | 0 | 0 | — |
case-16 | fail→pass | 7,830 | 3,956 | -49% | 1 | 1 | 0% | 1,165 | 1,084 | -7% | 0 | 0 | — |
case-17 | pass→pass | 4,498 | 1,713 | -62% | 1 | 1 | 0% | 742 | 648 | -13% | 0 | 0 | — |
case-18 | pass→pass | 3,904 | 1,569 | -60% | 1 | 1 | 0% | 604 | 566 | -6% | 0 | 0 | — |
case-19 | pass→pass | 11,239 | 5,339 | -52% | 1 | 1 | 0% | 1,875 | 1,287 | -31% | 0 | 0 | — |
case-20 | pass→pass | 4,109 | 3,934 | -4% | 1 | 1 | 0% | 680 | 1,037 | +53% | 0 | 0 | — |
case-21 | pass→fail | 3,884 | 4,288 | +10% | 1 | 1 | 0% | 787 | 1,332 | +69% | 0 | 0 | — |
case-22 | pass→pass | 9,789 | 8,089 | -17% | 1 | 1 | 0% | 1,929 | 1,781 | -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. 22 cases were attempted. The headline lift of +9 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.