icp-identification

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script (tam_builder.py) to research and build a Total Addressable Market. The instructions explicitly prioritize safety by requiring the use of --preview and --sample flags to show the user expected outcomes before any data is written to a database.
  • [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests and processes content from external company URLs during the research phase.
  • Ingestion points: External website content accessed via web research tools in Phase 1.
  • Boundary markers: None explicitly defined for the external data being processed.
  • Capability inventory: Includes execution of local scripts for market mapping and handoff to external lead-finding APIs (Apollo, Crustdata) and Supabase databases.
  • Sanitization: The skill mitigates risks by requiring the agent to synthesize a summary of the research and present it to the user for validation before proceeding to the ICP definition or lead execution phases.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 03:05 PM
Security Audit — agent-trust-hub — icp-identification