hotfix
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external information from bug reports which is subsequently used to generate documentation files and branch names, creating a potential surface for indirect injection.\n
- Ingestion points: User-provided bug descriptions and IDs are read in Phase 1 (SKILL.md).\n
- Boundary markers: There are no explicit delimiters defined to isolate the bug data from the agent's internal instruction set during interpolation.\n
- Capability inventory: The skill utilizes
Bashfor git operations,Write/Editfor recording hotfix details, andTaskfor coordinating with subagents.\n - Sanitization: No specific validation or sanitization steps are documented for the input strings before their use in paths or commands.\n- [SAFE]: The workflow incorporates a mandatory multi-agent approval process using the
Tasktool, requiring sign-offs from a lead programmer, QA tester, and producer before any deployment can proceed.\n- [SAFE]: Sensitive operations, such as writing to the production records directory, are explicitly gated by an instruction to request user permission before execution.
Audit Metadata