om-implement-feature
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
- [PROMPT_INJECTION]: The skill incorporates defensive instructions to identify and ignore common prompt injection patterns within untrusted content, requiring the agent to report such attempts rather than complying.
- [INDIRECT_PROMPT_INJECTION]: The skill manages an untrusted content boundary for processing feature briefs and repository data. It implements strict validation for external values and explicitly instructs the agent to treat all ingested content as data.
- [DYNAMIC_EXECUTION]: Support for repo-local extensions is provided through a controlled mechanism that prevents extensions from relaxing safety rules or expanding permissions.
- [COMMAND_EXECUTION]: The skill executes validation and testing commands within the scope of the project's worktree, with clear restrictions against exfiltration.
- [CREDENTIALS_UNSAFE]: A 'Secrets hygiene' policy is enforced, prohibiting the inclusion of credentials or sensitive data in any output or logs.
Audit Metadata