continue-codex-work
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary function is to manage session continuity by verifying session identity and rollout lineage. It uses a vendor-specific tool to read history and mandates a verification phase to ensure the context is valid before proceeding.
- [COMMAND_EXECUTION]: The instructions describe using standard shell techniques for persisting tasks across session interruptions, such as backgrounding processes with
nohuporsetsidand redirecting output to log files. These are documented as best practices for ensuring task resilience in terminal-based environments. - [EXTERNAL_DOWNLOADS]: The skill references official OpenAI documentation and version-pinned source code from the OpenAI GitHub repository to define its recovery logic. These are well-known technology service providers and repositories, and the references are used neutrally for technical guidance.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by ingesting data from session histories and project files (like
AGENTS.mdandCLAUDE.md). It mitigates this risk through a 'continuation contract' that requires independent verification of completed work and explicitly prohibits the agent from attempting to bypass safety or policy rejections recorded in the history.
Audit Metadata