メニュー
Codex CLI delegation model floors, sandboxes, and acceptance gates.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Cursor delegation model floors, modes, and acceptance gates.
Run Codex CLI as a subagent with GPT-5.6 at low or medium effort.
Parent-only orchestration: plan, route slices by capability floor, keep truth-judgment; never read or write application code.
Use the generic delegator workflow with the Cursor model mapping. Use when the user wants Cursor-based delegation, delegator-cursor, or the delegator workflow backed by Cursor subagents.
Use the generic delegator workflow with the Codex model mapping. Use when the user wants Codex-based delegation, delegator-codex, or the delegator workflow backed by Codex CLI/subagents.
Certainty-scored pull request reviews with reasoning and final ratings. Use when reviewing a PR, branch diff, or when the user asks for a PR review.
SOC 職業分類に基づく
| name | codex-model-mapping |
| description | Codex CLI delegation model floors, sandboxes, and acceptance gates. |
| disable-model-invocation | true |
Map generic delegation floors to Codex like this.
| Floor | Codex mapping | Sandbox | Gate |
|---|---|---|---|
| Scout | codex exec -m gpt-5.4-mini | workspace-write (scratch/findings only; do not patch application code) | Return scratch path, 3-line summary, confidence. Parent sanity-checks only. |
| Builder | codex exec -m gpt-5.6-terra with model_reasoning_effort="low" for small patches, medium when tests or logic are non-trivial | workspace-write | Relevant tests/build/checks pass or the delegatee reports the blocker. |
| Senior | codex exec -m gpt-5.5 with model_reasoning_effort="medium" | workspace-write; escalate sandbox only when the task genuinely needs network or broader access | Parent may inspect scratch notes and final diff before accepting. |
| You | Parent agent | Current host permissions | No delegation. |
Prefer the lowest floor that satisfies the slice. Go up one floor when stakes, reversibility, or ambiguity are unclear.
For naming be sure to prefix the agent name with the type of agent, like 'Scout:'.
