web-search

Pass

Audited by Gen Agent Trust Hub on Jun 2, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill implementation relies exclusively on the Python standard library (stdlib) for its core logic in scripts/web-search, which eliminates third-party dependency risks and minimizes the attack surface for supply chain vulnerabilities.
  • [SAFE]: API keys and sensitive credentials are handled appropriately using environment variables or configuration files in standard user directories (e.g., ~/.config/web-search/). The provided examples use clear placeholders and documentation includes explicit warnings against hardcoding secrets.
  • [SAFE]: Network communication is directed solely to established search provider APIs using secure HTTPS connections. The functionality for baseUrl overrides is correctly documented as a feature for corporate gateways, mirrors, and local development mocks.
  • [PROMPT_INJECTION]: As a tool designed to fetch external data, the skill exposes a surface for indirect prompt injection from search results ingested in scripts/web-search. This risk is mitigated by the skill's use of structured JSON output and distinct boundary markers (e.g., === provider === and --- WEBSEARCH ---) in its human-readable preview, which helps the agent separate search content from operational instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 2, 2026, 04:12 AM
Security Audit — agent-trust-hub — web-search