Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →ai-agent-camp用のCodexツーリングを設定するスキル。 「MCPサーバーを設定」「フックをインストール」「Codexの設定」「ツールセットアップ」「Codex CLIのインストール」等のリクエストで発動。
.claude/skills/minicoohei-aiagent-tooling-setup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -59% | 0% |
Codex CLI は OpenAI が提供するターミナルベースの AI コーディングアシスタントです。
Use this skill when the user needs Codex-specific tooling guidance.
docs/codex-mcp.md.bash scripts/install_hooks.shuv run python tools/check_command_paths.pyuv run python tools/credential_manager.py statusbashnpm install -g @openai/codex codex --version # verify installation
Requires Node.js 18+.
Set the OpenAI API key via environment variable:
bashexport OPENAI_API_KEY="your-api-key-here"
Or add OPENAI_API_KEY=... to the repo .env file (never commit this file).
| Setting | Value | Reason | |---------|-------|--------| | Sandbox | workspace-write | Restricts writes to the repo directory | | Approval | on-request | Asks before external commands |
Launch with:
bashcodex -a on-request
Avoid danger-full-access for normal learning flows. See docs/codex-safety.md.
npm install -g @openai/codex)OPENAI_API_KEYbash scripts/install_hooks.shaiagent-check-setup skill to verify environmentaiagent-lesson-runner skillAGENTS.md.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 11,043 | 5,963 | -46% | 1 | 1 | 0% | 1,878 | 1,467 | -22% | 0 | 0 | — |
case-02 | fail→pass | 11,641 | 8,208 | -29% | 1 | 1 | 0% | 1,990 | 1,390 | -30% | 0 | 0 | — |
case-03 | fail→pass | 17,610 | 7,153 | -59% | 1 | 1 | 0% | 2,940 | 1,682 | -43% | 0 | 0 | — |
case-04 | fail→pass | 6,888 | 2,235 | -68% | 1 | 1 | 0% | 1,082 | 733 | -32% | 0 | 0 | — |
case-05 | pass→pass | 7,995 | 1,734 | -78% | 1 | 1 | 0% | 1,196 | 698 | -42% | 0 | 0 | — |
case-06 | fail→pass | 14,897 | 3,298 | -78% | 1 | 1 | 0% | 2,230 | 904 | -59% | 0 | 0 | — |
case-07 | pass→pass | 12,444 | 2,696 | -78% | 1 | 1 | 0% | 1,715 | 854 | -50% | 0 | 0 | — |
case-08 | pass→pass | 10,894 | 18,287 | +68% | 1 | 1 | 0% | 1,586 | 687 | -57% | 0 | 0 | — |
case-09 | fail→pass | 12,191 | 5,376 | -56% | 1 | 1 | 0% | 2,030 | 1,444 | -29% | 0 | 0 | — |
case-10 | pass→pass | 13,249 | 4,905 | -63% | 1 | 1 | 0% | 1,987 | 1,151 | -42% | 0 | 0 | — |
case-11 | fail→pass | 7,206 | 2,691 | -63% | 1 | 1 | 0% | 1,070 | 757 | -29% | 0 | 0 | — |
case-12 | pass→pass | 12,023 | 3,661 | -70% | 1 | 1 | 0% | 1,661 | 1,000 | -40% | 0 | 0 | — |
case-13 | pass→pass | 3,057 | 1,777 | -42% | 1 | 1 | 0% | 412 | 688 | +67% | 0 | 0 | — |
case-14 | pass→pass | 17,935 | 3,867 | -78% | 1 | 1 | 0% | 931 | 757 | -19% | 0 | 0 | — |
case-15 | pass→pass | 10,038 | 3,456 | -66% | 1 | 1 | 0% | 1,589 | 1,015 | -36% | 0 | 0 | — |
case-16 | fail→pass | 9,244 | 1,982 | -79% | 1 | 1 | 0% | 1,411 | 760 | -46% | 0 | 0 | — |
case-17 | fail→pass | 10,847 | 3,686 | -66% | 1 | 1 | 0% | 1,750 | 834 | -52% | 0 | 0 | — |
case-18 | fail→pass | 7,313 | 2,048 | -72% | 1 | 1 | 0% | 1,280 | 739 | -42% | 0 | 0 | — |
case-19 | fail→pass | 15,128 | 3,710 | -75% | 1 | 1 | 0% | 2,576 | 992 | -61% | 0 | 0 | — |
case-20 | fail→pass | 3,796 | 1,595 | -58% | 1 | 1 | 0% | 533 | 669 | +26% | 0 | 0 | — |
case-21 | fail→pass | 8,183 | 2,213 | -73% | 1 | 1 | 0% | 1,230 | 764 | -38% | 0 | 0 | — |
case-22 | pass→pass | 7,050 | 4,121 | -42% | 1 | 1 | 0% | 1,191 | 1,128 | -5% | 0 | 0 | — |
case-23 | pass→pass | 7,868 | 3,768 | -52% | 1 | 1 | 0% | 1,310 | 1,108 | -15% | 0 | 0 | — |
case-24 | pass→pass | 3,553 | 5,788 | +63% | 1 | 1 | 0% | 477 | 949 | +99% | 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. 24 cases were attempted. The headline lift of +54 percentage points is the difference between those two pass rates over the 24 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.