grill-with-docs

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a validation step in the documentation that executes a local Node.js script: node skills/skill-authoring/scripts/validate-skill-library.mjs skills/grill-with-docs/SKILL.md. This script is part of the repository's internal validation framework.
  • [DATA_EXFILTRATION]: The skill is designed to read and explore the codebase and existing documentation (CONTEXT.md, ADRs) to verify user claims. While it accesses local project data, it lacks network capabilities to send this data externally.
  • [PROMPT_INJECTION]: The skill has a surface for Indirect Prompt Injection because it processes untrusted data from user plans and existing repository files.
  • Ingestion points: User-provided plans/designs, existing codebase files, and existing documentation (CONTEXT.md, ADRs).
  • Boundary markers: None explicitly defined in the instructions for separating ingested data from agent instructions.
  • Capability inventory: The skill uses tools to read the local filesystem (code exploration) and write/update markdown files (CONTEXT.md, ADRs).
  • Sanitization: No specific sanitization or escaping of external content is mentioned before the agent processes or writes it to documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 06:02 AM