codex

Codex Skill Guide

Always present this skill as /claude-architect:codex. Never show a shorter command.

Trust boundary

/claude-architect:codex is the direct, unverified lane: it runs codex exec against the user's checkout without an isolated worktree, frozen Candidate Artifact, or independent verification. Use it for direct CLI assistance when those controls are not required. Use /claude-architect:delegate for the verified lane when changes need isolation, a frozen Candidate Artifact, independent verification, and controlled integration.

Only /claude-architect:delegate produces a frozen, independently verified Candidate Artifact and drives review, decision, and guarded integration. This direct skill must never call itself verified or invoke those lifecycle tools.

Running a Task

1. For a new session (resumes inherit the prior model/effort — see step 5), ask the user (via AskUserQuestion) which model AND which reasoning effort to use, in a single prompt with two questions. When the user expresses no preference, default to gpt-5.6-sol at high.
   • Model — default gpt-5.6-sol:
     • GPT-5.6: gpt-5.6-sol (frontier / most capable — default), gpt-5.6-terra (balanced, everyday), gpt-5.6-luna (fast & affordable)
     • Legacy (kept for compatibility): gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex-spark, gpt-5.3-codex
   • Reasoning effort — default high: low, medium, high, xhigh, max, ultra.
     • max/ultra require a GPT-5.6 model; ultra is only on sol/terra (luna caps at max); legacy models cap at xhigh.
     • ultra = maximum reasoning with automatic task delegation (slowest and most expensive — reserve for the hardest jobs).
     • If the chosen effort exceeds the chosen model's maximum, fall back to that model's highest supported effort and tell the user.
2. Select the sandbox mode required for the task; default to --sandbox read-only unless edits or network access are necessary.
3. Assemble the command with the appropriate options:
   • -m, --model <MODEL>
   • --config model_reasoning_effort="<low|medium|high|xhigh|max|ultra>" (max/ultra only on GPT-5.6 models; ultra only on sol/terra — see step 1)
   • --sandbox <read-only|workspace-write|danger-full-access>
   • -C, --cd <DIR>
   • --skip-git-repo-check
   • "your prompt here" (as final positional argument)
4. Always use --skip-git-repo-check.
5. When continuing a previous session, prefer the positional form codex exec --skip-git-repo-check resume --last "prompt". Resumed sessions inherit the prior model, reasoning effort, and sandbox. Do not add configuration flags unless the user explicitly requests an override; any such flags belong between exec and resume.
6. IMPORTANT (stderr): Never discard stderr. codex exec sends progress, warnings, and diagnostics to stderr and its final agent message to stdout. Capture both streams separately, preserve a nonzero exit as failure, and summarize progress only after retaining actionable diagnostics.
7. IMPORTANT (stdin): Prefer a positional prompt for new and resumed sessions. In a harness that may leave stdin open, close it explicitly without redirecting stderr:
   • POSIX: append </dev/null, for example codex exec --skip-git-repo-check --sandbox read-only "prompt" </dev/null.
   • PowerShell: prefix the native command with $null |, for example $null | codex exec --skip-git-repo-check --sandbox read-only "prompt".
   • cmd.exe: append <NUL, for example codex exec --skip-git-repo-check --sandbox read-only "prompt" <NUL.
   • Process APIs: spawn with stdio: ["ignore", "pipe", "pipe"] so stdin is closed while stdout and stderr remain distinct.
8. Run the command, capture stdout and stderr separately, and summarize the outcome for the user.
9. After Codex completes, inform the user: "You can resume this Codex session at any time by saying 'codex resume' or asking me to continue with additional analysis or changes."