run-codex-subagents

Pass

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the installation of the codex-worker package from the public NPM registry using npm install -g codex-worker or via npx -y codex-worker for ad-hoc execution.
  • [COMMAND_EXECUTION]: Executes the codex-worker CLI to orchestrate coding agents, which includes starting threads, managing execution turns, and interacting with a local background daemon.
  • [PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface through its design of reading and processing external Markdown files for agent instructions.
  • Ingestion points: Markdown prompt files such as task.md, mission.md, and followup.md are ingested via codex-worker run and send commands.
  • Boundary markers: Mission templates utilize structured headers like ## Mission, ## Constraints, and ## Definition Of Done to delineate agent instructions.
  • Capability inventory: Orchestrated coding agents possess capabilities to execute shell commands (e.g., npm test, npm run build) and perform file system modifications within the project workspace.
  • Sanitization: There is no documented evidence of input validation or sanitization for the content within the Markdown prompt files processed by the agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 20, 2026, 11:41 PM