Agent skill
codex-cli-subagent-transport
Use only when native subagent execution is unavailable and the caller needs one deterministic Codex CLI run with repo-local artifacts and a manifest-based result contract.
Install this agent skill to your Project
npx add-skill https://github.com/shaowei-g/codex-skills/tree/main/skills/codex-cli-subagent-transport
SKILL.md
Codex CLI Subagent Transport
Use this skill only for delegated transport. It does not own routing, phase judgment, acceptance, repair, or specialist semantics.
Load First
./references/manifest-based-run-contract.md
Authority Boundary
- This skill owns only the mechanics of one Codex CLI execution.
- The caller owns prompt content, delegated scope, response validation, acceptance, and retry policy.
- This skill must not infer feature state, phase readiness, or acceptance from logs.
Required Inputs
Provide these values explicitly:
- repo root
- feature slug
- assigned phase
- assigned subagent
- prompt file
- model when non-default
- run id when the caller wants deterministic naming
- run root when the caller overrides the default repository-local run location
Required Outputs
Every run must materialize:
- repo-local run directory
prompt.mdresponse.mdexec.logmanifest.envmanifest.json- printed path assignments for the caller
Invocation
Preferred invocation:
bash ./scripts/run_codex_cli_subagent.sh \
--repo-root /abs/path/to/repo \
--feature <feature-slug> \
--phase <assigned-phase> \
--subagent <assigned-subagent> \
--prompt-file /abs/path/to/prompt.md
Stop Rule
- Run exactly one
codex execcall. - Always emit manifest files and printed path assignments after the run finishes.
- Do not retry, reroute, or reinterpret the result inside this skill.
- Treat
response.mdas the only authoritative payload andexec.logas diagnostics only.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
ekoen-frontend-code-review
Production-grade code review for the eKoEN frontend codebase. Use when asked to review, audit, inspect, or harden eKoEN React/Next.js frontend code, pages, components, hooks, API clients, or state layers for bugs, performance bottlenecks, TypeScript safety issues, request inefficiency, state-management risks, memory leaks, rendering regressions, error-handling gaps, duplication, or maintainability problems, especially when a structured severity-ranked report with code locations, root causes, suggested fixes, and example refactors is required.
ekoen-backend.doc.route-error
Add HttpErrorRes responses to eKoEN backend routes when controllers can throw errors from app/util/errors/index.ts
conventional-commit-helper
Generate and validate Conventional Commit messages for semantic-release workflows. Use when a user asks to write a commit message, asks to commit changes, asks for commit type/scope selection, or asks to enforce Conventional Commits consistency.
skill-usage-logger
All skill MUST use this logger to record usage events in NDJSON format.Log Codex skill usage events to NDJSON with safe concurrent appends.
nodejs-package-json
Patch a Node.js project's package.json for standard build automation. Use when asked to add/standardize package.json scripts like prebuild/postbuild for TypeScript builds (rimraf dist + tsc-alias), adjust prebuild for Next.js (rimraf dist .next), or ensure an existing package.json pkg config includes required scripts/assets/targets/outputPath.
Didn't find tool you were looking for?