workflow-grilling
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides instructional guidelines for a conversational planning workflow. No malicious code, obfuscation, or unauthorized access attempts were found within the instructions.
- [EXTERNAL_DOWNLOADS]: The skill references a public GitHub repository (github.com/mattpocock/skills) for attribution purposes. This link targets a well-known source and does not involve automated downloads or runtime execution of remote code.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided plans and decisions, which constitutes an attack surface for untrusted data. However, the workflow requires one-question-at-a-time interaction and manual confirmation for each step, providing a natural human-in-the-loop mitigation. Ingestion points: User responses to interview questions (SKILL.md). Boundary markers: Present as explicit 'Confirm' and 'Wait' steps after each recommendation. Capability inventory: The instructions mention the agent's ability to look up facts via filesystem access, git history, and command execution. Sanitization: None specified.
Audit Metadata