audience-mapper
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection through its ingestion of untrusted external data such as CRM exports, social media insights, and surveys. Ingestion points: External data sources and files referenced in SKILL.md. Boundary markers: The skill explicitly instructs the agent to treat fetched files as untrusted and to never follow instructions embedded in CSVs or social posts. Capability inventory: Capabilities are limited to local file writes to the memory/ directory; the skill lacks tools for network operations, command execution, or dynamic code evaluation. Sanitization: Relies on clear instructional guardrails and a security boundary defined in SKILL.md.
- [SAFE]: No malicious code, obfuscated content, unauthorized persistence mechanisms, or risky remote dependencies were detected in the skill files.
Audit Metadata