knowledge-curator

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted inputs including career knowledge proposals and evidence. While it follows a governed review process, the integration of this data into the agent's context provides a surface for indirect prompt injection attacks designed to influence knowledge curation decisions.\n
  • Ingestion points: Processes "career knowledge candidates" and "proposals" within SKILL.md.\n
  • Capability inventory: The skill can trigger canonical knowledge updates through an admission tool and by passing data to the output-career-portfolio skill.\n
  • Boundary markers: The skill relies on authority verification and policy checks but lacks explicit delimiters for untrusted content.\n
  • Sanitization: The skill uses a structured YAML schema for curation records but does not specify input sanitization or escaping techniques for user-provided strings.\n- [EXTERNAL_DOWNLOADS]: The skill performs a periodic version check for an optional update tool when network access is available.\n
  • Evidence: SKILL.md instructs the agent to check for updates to the why-hire-me-update skill using a cached network request.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 09:16 PM
Security Audit — agent-trust-hub — knowledge-curator