Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →ai-agent-campの環境変数・認証情報を安全に管理するスキル。 「APIキーを設定」「.envを管理」「環境変数の設定」「credential managerを使いたい」「秘密情報の管理」等のリクエストで発動。
.claude/skills/minicoohei-aiagent-env-manager/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -28% | 0% |
Use this skill for safe environment setup.
.env.example.uv run python tools/credential_manager.py store <KEY> over editing secrets into markdown.uv run python tools/credential_manager.py status to verify setup without printing values..env.exampletools/credential_manager.pydocs/codex-safety.md.env management or OS credential storage.uv run python tools/credential_manager.py store <KEY> when possible..env is required, limit guidance to key names and file locations.uv run python tools/credential_manager.py status or masked file checks..env is local-only and must stay out of git.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 10,742 | 5,474 | -49% | 1 | 1 | 0% | 1,978 | 1,081 | -45% | 0 | 0 | — |
case-02 | fail→pass | 15,588 | 3,471 | -78% | 1 | 1 | 0% | 2,797 | 732 | -74% | 0 | 0 | — |
case-03 | fail→fail | 2,119 | 4,803 | +127% | 1 | 1 | 0% | 277 | 982 | +255% | 0 | 0 | — |
case-04 | fail→fail | 10,723 | 5,737 | -46% | 1 | 1 | 0% | 1,746 | 1,236 | -29% | 0 | 0 | — |
case-05 | fail→fail | 15,102 | 7,543 | -50% | 1 | 1 | 0% | 2,766 | 1,593 | -42% | 0 | 0 | — |
case-06 | pass→pass | 9,896 | 6,189 | -37% | 1 | 1 | 0% | 1,576 | 1,206 | -23% | 0 | 0 | — |
case-07 | fail→pass | 10,888 | 3,630 | -67% | 1 | 1 | 0% | 1,792 | 777 | -57% | 0 | 0 | — |
case-08 | fail→pass | 4,483 | 4,719 | +5% | 1 | 1 | 0% | 593 | 631 | +6% | 0 | 0 | — |
case-09 | pass→pass | 4,244 | 1,750 | -59% | 1 | 1 | 0% | 570 | 428 | -25% | 0 | 0 | — |
case-10 | fail→pass | 9,316 | 4,660 | -50% | 1 | 1 | 0% | 1,394 | 1,000 | -28% | 0 | 0 | — |
case-11 | pass→pass | 9,889 | 4,927 | -50% | 1 | 1 | 0% | 1,659 | 1,022 | -38% | 0 | 0 | — |
case-12 | pass→pass | 16,887 | 9,901 | -41% | 1 | 1 | 0% | 2,678 | 1,550 | -42% | 0 | 0 | — |
case-13 | fail→pass | 6,188 | 2,587 | -58% | 1 | 1 | 0% | 806 | 539 | -33% | 0 | 0 | — |
case-14 | fail→pass | 10,751 | 6,857 | -36% | 1 | 1 | 0% | 1,877 | 550 | -71% | 0 | 0 | — |
case-15 | fail→pass | 7,094 | 3,287 | -54% | 1 | 1 | 0% | 1,186 | 572 | -52% | 0 | 0 | — |
case-20 | pass→pass | 6,816 | 4,906 | -28% | 1 | 1 | 0% | 1,222 | 1,094 | -10% | 0 | 0 | — |
case-16 | fail→pass | 18,494 | 2,068 | -89% | 1 | 1 | 0% | 450 | 493 | +10% | 0 | 0 | — |
case-17 | pass→pass | 4,548 | 7,787 | +71% | 1 | 1 | 0% | 649 | 672 | +4% | 0 | 0 | — |
case-18 | fail→pass | 6,077 | 2,076 | -66% | 1 | 1 | 0% | 828 | 526 | -36% | 0 | 0 | — |
case-19 | fail→pass | 13,063 | 2,674 | -80% | 1 | 1 | 0% | 1,799 | 673 | -63% | 0 | 0 | — |
case-21 | pass→pass | 7,601 | 3,916 | -48% | 1 | 1 | 0% | 1,252 | 852 | -32% | 0 | 0 | — |
case-22 | pass→pass | 15,734 | 3,295 | -79% | 1 | 1 | 0% | 636 | 816 | +28% | 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 +50 percentage points is the difference between those two pass rates over the 22 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.