ai-consultant

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill framework establishes an attack surface for indirect prompt injection by design, as it relies on the ingestion of untrusted data from external sources to generate consulting deliverables.\n
  • Ingestion points: Untrusted data enters the agent context through web searches performed in Phase 1 (Pre-Engagement Research) and stakeholder interview notes collected in Phase 3 (Stakeholder Discovery) as described in SKILL.md.\n
  • Boundary markers: Absent. The templates provided in references/deliverable-templates.md do not include delimiters or specific instructions for the agent to disregard potential instructions embedded within the gathered research or interview notes.\n
  • Capability inventory: The skill utilizes web_search and file system write operations to generate documents such as company-briefing.md, discovery-report.md, and project proposals.\n
  • Sanitization: Absent. The skill instructions do not include steps to validate, filter, or sanitize content from external sources before it is interpolated into the final deliverables.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 05:26 PM
Security Audit — agent-trust-hub — ai-consultant