ask-matt

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines and documents workflows that process untrusted data from external sources, presenting a risk that adversarial content could influence agent behavior.
  • Ingestion points: Mentions processing raw bug reports and feature requests in the triage flow, primary sources in the research flow, and human feedback in the questionnaire flow (SKILL.md).
  • Boundary markers: Incorporates phase boundaries and explicit context clearing (/clear) to manage information flow between tasks (SKILL.md, PHASE-BOUNDARIES.md).
  • Capability inventory: The router leads to high-privilege operations including code implementation, test execution, and script generation (/implement, /tdd, /wizard).
  • Sanitization: Instructs the agent to reconcile work capsules against existing repository state and tests before accepting results (/resume-work in SKILL.md).
  • [DYNAMIC_EXECUTION]: The skill documentation describes the /wizard tool, which dynamically generates interactive bash scripts to assist with provisioning, credential setup, and secret management (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 12:02 AM
Security Audit — agent-trust-hub — ask-matt