Codex
Profiles
- (no profile) — read-only analysis (no compilation, no tests)
- — local edits, build, or test (, full-auto)
- — network or broad filesystem access (, full-auto)
Use
whenever the task involves
,
, or any command that writes build artifacts. The default read-only sandbox blocks all writes including compiler temp dirs.
For complex tasks, add
--config model_reasoning_effort="xhigh"
.
Model Selection
Default to the project model in
(
).
Use
--model gpt-5.3-codex-spark
for speed-first interactive work:
- tight back-and-forth iteration
- quick bug fixes and targeted edits
- small-to-medium refactors or UI tweaks
- rapid prototyping where frequent redirect/interrupt is expected
Stay on default
for depth-first work:
- long autonomous runs
- broad multi-file changes with heavy reasoning
- test-heavy or verification-heavy tasks
- higher correctness/risk surfaces where deeper reasoning is worth the extra latency
If the user explicitly requests a model, follow the user's request.
Commands
Project defaults in
. Always include
and
. Only override model when the user explicitly requests it, or when the task clearly fits the Spark speed-first criteria above.
| Use case | Command |
|---|
| Read-only analysis | codex exec --skip-git-repo-check -o "$CODEX_OUT" "prompt" 2>/dev/null
|
| Speed-first interactive loop (Spark) | add --model gpt-5.3-codex-spark
|
| Apply local edits | add |
| Network / broad access | add |
| Resume session | echo "prompt" | codex exec --skip-git-repo-check -o "$CODEX_OUT" resume --last 2>/dev/null
|
Create a temp file for output before running:
CODEX_OUT=$(mktemp /tmp/codex-out.XXXXXX)
. Read the result from
after completion.
Run with Bash
. Monitor via
with
block: true, timeout: 600000
— do NOT use
commands to poll. Read the
file for the final answer.
Resuming
Don't pass model/effort/sandbox flags when resuming — original session settings carry over.
is safe to include.
Conventions
This is a Go project. Use
,
,
in prompts. Codex workers don't receive CLAUDE.md or hooks, so project conventions must be explicit.
Post-Execution Verification
Always verify Codex output via — never trust the worker's self-reported file list. Codex workers may make out-of-scope changes (deleting files, reverting unrelated code, removing test functions). The completion promise may list only intended files while omitting destructive side effects.
When other work exists in the same codebase, include explicit "DO NOT modify or delete" file lists in the prompt. The positive instruction ("only touch these files") is insufficient — Codex interprets "clean up" broadly.
Per [[principles/prove-it-works]]: trust artifacts, not self-reports.
Worktree Usage
With
, the worktree has its own git index.
may symlink to
— use
paths for git operations.