implement
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied specifications or tickets, which represents an attack surface for instructions embedded in data.
- Ingestion points: External task descriptions (spec or tickets) are ingested into the agent context in SKILL.md.
- Boundary markers: No delimiters or instructions are used to distinguish user data from the skill's instructions.
- Capability inventory: The agent can run shell commands for typechecking/testing and write to the filesystem via git commits as instructed in SKILL.md.
- Sanitization: No sanitization or validation logic is applied to the ingested content.
Audit Metadata