grill-designs
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external sources such as repository files, documentation, and command outputs to resolve design decisions automatically. This creates a vulnerability surface where an attacker could place malicious instructions in a repository file that the agent then adopts as truth or follows during the 'grilling' process.
- Ingestion points: Processes repository files, documentation, tests, and outputs from system commands (Step 1 of the Loop).
- Boundary markers: The instructions do not define boundary markers or explicit instructions for the agent to ignore potentially malicious content embedded in the ingested data.
- Capability inventory: The skill uses general command execution and file reading to gather context, and later triggers the
write-planstool. - Sanitization: There are no instructions for sanitizing or validating the data retrieved from external files before it is incorporated into the decision tree.
Audit Metadata