open-websearch

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines clear procedures for environment discovery and setup using shell commands like 'open-websearch serve' and 'open-websearch status'. Security is maintained by requiring the agent to obtain user confirmation before performing these actions.
  • [EXTERNAL_DOWNLOADS]: Installation paths for dependencies such as 'playwright' and the core 'open-websearch' tool are documented using standard package managers (npm/npx). The instructions emphasize collecting prerequisites and user consent before initiating any downloads.
  • [PROMPT_INJECTION]: A dedicated safety section addresses the risk of indirect prompt injection from external web data. 1. Ingestion points: Web content retrieved via 'search' and 'fetchWebContent' tools as defined in 'references/tools.md'. 2. Boundary markers: Explicit instructions in 'SKILL.md' to treat all fetched results as untrusted content. 3. Capability inventory: Subprocess management for daemon control and package installation. 4. Sanitization: Mandatory logic to ignore instructions within fetched content and prevent execution of code snippets found on web pages.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:36 PM
Security Audit — agent-trust-hub — open-websearch