closing-obsolete-issues
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub issue descriptions and comments. This creates a surface for indirect prompt injection where an attacker could embed instructions in an issue to influence the agent's analysis or comment generation.
- Ingestion points: External content is ingested via the
gh issue viewcommand as described inSKILL.md. - Boundary markers: The instructions do not explicitly define boundary markers or delimiters for the ingested issue content.
- Capability inventory: The skill has the capability to close issues via
gh issue closeand modify the localreferences/rationale_templates.mdfile. - Sanitization: No specific sanitization or filtering of the ingested issue text is mentioned.
- Mitigation: The risk is effectively mitigated by the 'User Approval Required' mandate in Instruction 3, which requires the agent to present all proposed actions and comments to a human for explicit review before execution.
Audit Metadata