efficient-web-research
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No malicious instruction overrides, role-play jailbreaks, or safety filter bypass attempts were detected in the skill instructions.
- [DATA_EXFILTRATION]: The skill defines protocols for fetching external data via GitHub API,
read_url_content, andbrowser_subagent. These are standard tools for web research and there are no patterns indicating the exfiltration of sensitive local environment data or credentials. - [REMOTE_CODE_EXECUTION]: No patterns involving the execution of remote scripts (e.g.,
curl | bash) or installation of untrusted packages were found. - [INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data from external websites and GitHub repositories.
- Ingestion points: Data enters the agent context via
read_url_content,browser_subagent, and GitHub API responses. - Boundary markers: The skill lacks explicit instructions to treat fetched content as untrusted or to ignore instructions embedded within the fetched text (e.g., HTML comments or hidden markdown).
- Capability inventory: The skill uses web search and content reading tools but does not explicitly request file system write access or shell execution.
- Sanitization: It recommends stripping boilerplate (navigation, ads, footers) to save tokens, which provides minimal sanitization but does not protect against adversarial content. This represents a typical attack surface for research-oriented skills.
Audit Metadata