advisor
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill aggregates user requests and system outputs to provide context to an advisor sub-agent.
- Ingestion points: User requests verbatim, test outputs, and git diffs are compiled via the briefing template.
- Boundary markers: Employs markdown headers and specific sections in
references/briefing-template.mdto delimit input types. - Capability inventory: Utilizes the
Tasktool and sub-agent spawning with IDs to manage consultation sessions. - Sanitization: Includes mandatory instructions to redact API tokens, keys, and environment variables, and requires conversation ID validation before sharing transcripts.
- [SAFE]: The skill follows secure state management practices by instructing the agent to avoid committing state files to version control and preferring managed file tools over shell redirects.
Audit Metadata