implement
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external PRDs and issue descriptions, which represent untrusted ingestion points for the agent.
- Ingestion points: The agent reads PRDs and project issues from the local environment to drive implementation.
- Boundary markers: The instructions lack delimiters or explicit directives to ignore potential instructions embedded within the ingested documents, creating a vulnerability to indirect injection.
- Capability inventory: The skill grants the agent authority to run shell commands (typechecking and test suites) and perform file system writes (committing to the current branch).
- Sanitization: No input validation or content filtering is specified in the skill instructions to mitigate the risk of malicious commands being parsed from the input documentation.
Audit Metadata