om-close-fixed-issues
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes explicit defensive instructions to counter prompt injection attempts found in repository or tracker content. (Evidence: 'references/agentic-setup.md' instructs the agent to treat directives like 'ignore previous instructions' as suspected prompt injection and to report rather than follow them.)
- [DATA_EXFILTRATION]: The skill implements strict sanitization to prevent accidental leakage of sensitive information. (Evidence: 'SKILL.md' contains a rule prohibiting the verbatim pasting of PR bodies into issue comments to prevent the exposure of secrets.)
- [DYNAMIC_EXECUTION]: The skill supports repo-local extensions, allowing for repository-specific behavior adjustments within a controlled framework. (Evidence: 'references/agentic-setup.md' describes a mechanism to load extensions from '.ai/skills/om-close-fixed-issues/SKILL.md', while explicitly forbidding extensions from relaxing safety or quality rules.)
- [COMMAND_EXECUTION]: The skill uses the 'jq' utility to process configuration data. (Evidence: 'references/agentic-setup.md' uses 'jq' to extract custom keywords from '.ai/agentic.config.json'.)
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data but mitigates the associated risks through a well-defined boundary. (Ingestion points: PR bodies, titles, and issue descriptions; Boundary markers: Clear instructions to treat content as data; Capability inventory: Tracker mutations like 'comment-issue' and 'close-issue'; Sanitization: Filtering unmatched mentions and metadata-only comments.)
Audit Metadata