grill-me
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as Project Requirements Documents (PRDs), architecture drafts, and product ideas to build a decision tree for questioning. If these documents contain malicious instructions, the agent might follow them, potentially leveraging its capabilities to read the local environment.
- Ingestion points: Untrusted data enters the agent context via user-provided PRDs, architecture drafts, migration plans, and product ideas (referenced in
SKILL.md). - Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the processed data.
- Capability inventory: The skill is instructed to check facts by reading the local repository, configuration files, logs, and documentation (
SKILL.md). It also acknowledges the ability to run commands and write to the system, subject to host permissions. - Sanitization: No evidence of sanitization, escaping, or validation of the ingested external content was found in the instructions.
Audit Metadata