implement
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources described as "spec or tickets" (SKILL.md).
- [INDIRECT_PROMPT_INJECTION]: The skill maintains a capability inventory including shell command execution for typechecking and testing, as well as file system operations for committing work to a repository (SKILL.md).
- [INDIRECT_PROMPT_INJECTION]: There is a lack of explicit boundary markers or instructions to ignore embedded commands within the processed specifications, increasing the risk that malicious instructions in the input could be followed by the agent.
- [INDIRECT_PROMPT_INJECTION]: No sanitization or validation mechanisms are defined for the external content before it is processed or used to influence agent actions.
Audit Metadata