work-operating-model

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests historical context from the environment's memory tool to inform the interview process. \n
  • Ingestion points: Phase 2 of SKILL.md uses the search_thoughts tool to retrieve historical data. \n
  • Boundary markers: Instructions explicitly command the agent to treat search results as 'hints, not facts' and 'tentative context'. \n
  • Capability inventory: The skill possesses write capabilities through the save_operating_model_layer and capture_thought tools. \n
  • Sanitization: The workflow requires mandatory user confirmation of summaries and checkpoints before any data is persisted. \n- [NO_CODE]: The skill package consists solely of instructional markdown and metadata configuration, with no bundled executable scripts or automated installation code.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 08:05 AM
Security Audit — agent-trust-hub — work-operating-model