having-difficult-conversations
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFENO_CODE
Full Analysis
- NO_CODE (SAFE): No executable code, shell scripts, or external dependencies are included in the skill. All content is static documentation or templates.
- DATA_EXFILTRATION (SAFE): No network requests (curl, wget) or file access commands for sensitive system locations are present.
- PROMPT_INJECTION (SAFE): No patterns aimed at overriding safety filters or extracting system instructions were detected.
- INDIRECT_PROMPT_INJECTION (SAFE): The skill ingests user context but lacks tool capabilities to execute data-driven attacks. 1. Ingestion points: User input via references/INTAKE.md. 2. Boundary markers: Instruction in SKILL.md to use anonymized summaries and avoid secrets. 3. Capability inventory: Limited to generating Markdown text. 4. Sanitization: Relies on user-led anonymization.
Audit Metadata