Shapes turns action-first: action leads, steps numbered, state restated. Use for ADHD-friendly output. Do not use to trim tokens; use response-compression.
The English summary is being prepared. Showing the original description.
Delegates tasks to the OpenAI Codex CLI (codex exec) via delegation-core. Use when delegation-core selects Codex or OpenAI credentials are already configured.
The English summary is being prepared. Showing the original description.
Review the instructions the agent will receive before you install.
OpenAI's Codex CLI ships as npm @openai/codex and installs a codex
binary. Its headless entry point is codex exec, which takes the
prompt as a positional argument and reuses saved CLI credentials.
delegation-core selected codex for the taskcodex login has never succeeded there.
See the login caveat belownpm install -g @openai/codex
Homebrew and a curl installer are also documented upstream.
Either persist an API key or log in interactively:
printenv OPENAI_API_KEY | codex login --with-api-key
Verification runs codex login status, which reports one of "Logged
in using an API key", "Logged in using ChatGPT", "Logged in using
Agent Identity", or "Not logged in". The service uses that probe
rather than checking a single variable, because either credential
path is valid and only the probe sees both.
CODEX_API_KEY scopes a credential to a single codex exec run.
uv run python scripts/delegation_executor.py codex "Explain this module" \
--files src/
make -C plugins/conjure delegate-codex PROMPT='Explain this module'
codex exec "Explain the retry logic in this package"
codex declares the code_execution strength and ranks behind gemini,
qwen, minimax, glm and muse in the candidate order. It declares no
model ids, so smart_delegate leaves model selection to the CLI.
| Property | Value |
|---|---|
| Binary | codex |
| Package | @openai/codex |
| Headless form | codex exec <prompt> |
| Prompt delivery | positional, stdin also supported |
| Version probe | codex --version |
| Auth probe | codex login status |
| Structured output | codex exec --json |
Login caveat worth knowing before scripting this: an open issue
(openai/codex#9253) reports that codex login fails outright on
headless hosts, over SSH or in containers, unless a workspace
administrator has separately enabled device-code auth. That is a
single reporter's account with no maintainer confirmation, but it is
the failure mode to check first when codex exec will not run
unattended. Persisting an API key beforehand avoids it.
codex resolves on PATH and answers codex --versioncodex login status reports a logged-in state rather than "Not
logged in"codex exec <prompt> with the prompt last
and no prompt flag before itdelegations/codex/YYYYMMDD_HHMMSS.mddelegation-core selected codexNo reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Shapes turns action-first: action leads, steps numbered, state restated. Use for ADHD-friendly output. Do not use to trim tokens; use response-compression.
The English summary is being prepared. Showing the original description.
Inverts burden of proof for code additions. Use when reviewing PRs, planning refactors, or running unbloat to challenge every addition's necessity.
The English summary is being prepared. Showing the original description.
Tracks per-agent token usage and flags waste in parallel dispatch. Use when evaluating parallel agent efficiency or after a multi-agent run.
The English summary is being prepared. Showing the original description.
Coordinates Claude agent teams via filesystem protocol. Use when orchestrating parallel agents with task dependencies. Do not use for single-agent tasks.
The English summary is being prepared. Showing the original description.
Evaluates API surface design, consistency, and exemplar alignment. Use when reviewing public API changes or before releasing a new API surface.
The English summary is being prepared. Showing the original description.
Selects architecture paradigm via research before scaffolding. Use when architecture is undecided and the choice needs justification and documentation.
The English summary is being prepared. Showing the original description.