grill-with-docs
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs expected knowledge management tasks, including reading and updating domain context files (glossaries) and Architecture Decision Records (ADRs) within the local workspace.
- [COMMAND_EXECUTION]: The instructions mention standard development tools such as
git status,bun test, and a project-specificwikiCLI for configuration management. These are routine operations for the described development environment. - [DATA_EXPOSURE]: The skill accesses local documentation files (
CONTEXT.md,ADRfiles) to perform its core function of domain language validation. Access is scoped to the workspace directory and follows the intended purpose of architecture review. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documentation files (ADRs and glossaries) as input. While this represents an ingestion surface, the instructions specifically direct the agent to 'interview' the user and 'challenge' terms, which naturally limits the execution of potentially malicious instructions embedded in those documents.
Audit Metadata