browsing-history

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's primary function involves reading local Chrome history data from the standard application support path and storing it in a local SQLite database at ~/data/browsing.db. This behavior is consistent with the skill's stated purpose.
  • [COMMAND_EXECUTION]: The browsing_query.py script uses subprocess.run to execute the 'llm' CLI tool for its optional categorization feature. This is a documented dependency and the execution uses a hardcoded path with arguments passed as a list, which is the secure way to handle subprocess calls.
  • [PROMPT_INJECTION]: The skill processes website titles and URLs from the user's history to generate content categories, creating an indirect prompt injection surface.
  • Ingestion points: Browsing history data (URLs and titles) enters the context through the local SQLite database.
  • Boundary markers: The data is interpolated into a prompt for the 'llm' tool without explicit delimiters or 'ignore' instructions.
  • Capability inventory: The skill can execute the 'llm' CLI tool via subprocess.
  • Sanitization: Website titles are truncated but not otherwise sanitized. This is evaluated as low risk because the LLM output is exclusively used for categorization labels in the user's report.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 10:20 PM
Security Audit — agent-trust-hub — browsing-history