competitor-teardown

Pass

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references installation instructions for the 'belt' CLI from a GitHub repository and uses 'npx' to add the 'belt-sh/cli' package to the environment.
  • [COMMAND_EXECUTION]: Instructions guide the user to execute the 'belt' command-line tool for authentication, research tasks, and running various analytical applications.
  • [REMOTE_CODE_EXECUTION]: The skill uses a remote 'python-executor' tool to execute Python code for generating positioning maps, involving the use of the matplotlib library.
  • [DATA_EXFILTRATION]: The skill is designed to extract text data and capture screenshots from external competitor URLs for analysis purposes.
  • [PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by ingesting untrusted content from external websites into the agent's context.
  • Ingestion points: External URLs processed by 'tavily/extract' and 'infsh/agent-browser' tools.
  • Boundary markers: No delimiters or specific instructions to ignore embedded commands are present in the provided templates.
  • Capability inventory: Includes remote code execution via 'python-executor', file system interaction for saving images, and access to web research tools.
  • Sanitization: Content extraction is performed for analysis without explicit sanitization or filtering instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 8, 2026, 08:29 PM
Security Audit — agent-trust-hub — competitor-teardown