whychk
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze decision text which may originate from untrusted sources or contain embedded instructions intended to manipulate the agent.
- Ingestion points: User-provided decision summaries processed in Step 1 of the workflow in
SKILL.md. - Boundary markers: The skill lacks explicit instructions to treat the 'decision' input as data only or to use delimiters to prevent instruction leakage.
- Capability inventory: The skill instructs the agent to write flags to plans, issues, or commit messages (Step 8), creating a write-back path for potentially tainted analysis.
- Sanitization: No input validation or sanitization of the decision text is specified before it is passed to the critic sub-session.
Audit Metadata