won-deal-icp-finder

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local Python script (scripts/analyze.py) to process deal data. This is used for calculating revenue aggregates, ranking deals, and segmenting data by industry and size.
  • [DYNAMIC_EXECUTION]: The skill uses python3 to run its internal analysis engine. The script is part of the skill package and performs deterministic arithmetic on the provided dataset.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted CRM data from HubSpot or CSV exports, which serves as a potential injection surface for data-driven attacks.
  • Ingestion points: Processes external deal and company data via the HubSpot MCP or local files (e.g., /tmp/deals.json).
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are applied when interpolating data into the archetype generation or the sales-nav-search-builder prompt.
  • Capability inventory: The skill has the ability to execute local Python scripts and trigger sibling skills via sendPrompt commands.
  • Sanitization: The scripts/analyze.py utility provides extensive normalization and validation for currency formats, dates, and numeric company sizes, which provides a layer of protection against malformed input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:42 AM
Security Audit — agent-trust-hub — won-deal-icp-finder