om-auto-implement-issue
Pass
Audited by Gen Agent Trust Hub on Jul 18, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes a dedicated 'Untrusted content boundary' section that explicitly instructs the agent to treat all repository and tracker content (issue bodies, PR comments, etc.) as data to analyze rather than instructions to obey. It specifically identifies and forbids compliance with common injection patterns like 'ignore previous instructions'.
- [DATA_EXFILTRATION]: Strict prohibitions are in place to prevent the exfiltration of credentials or sensitive data into PR comments or external logs. The skill restricts tool use to the repository context and its authorized tracker.
- [COMMAND_EXECUTION]: The skill performs shell commands for project building and testing but implements safety checks. All externally sourced values, such as issue IDs and branch slugs, are validated using numeric or alphanumeric regex patterns before being interpolated into shell commands or file paths.
- [PROMPT_INJECTION]: The skill defines a protocol for repository-local extensions (local SKILL.md files) but explicitly mandates that these local rules cannot relax the primary skill's safety/quality constraints or expand its tool/network access.
Audit Metadata