efficient-web-research
Pass
Audited by Gen Agent Trust Hub on Aug 1, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate instructional guidelines for improving the efficiency of an AI agent's web research tasks. It uses standard tools such as
read_url_content,search_web, and official GitHub APIs to perform its functions. No evidence of obfuscation, hardcoded credentials, malicious persistence, or unauthorized data exfiltration was found.\n- [PROMPT_INJECTION]: The skill operates by ingesting content from untrusted external sources, creating a surface for indirect prompt injection.\n - Ingestion points: External data enters the agent context via GitHub API calls, non-GitHub URLs (processed through
read_url_contentandbrowser_subagent), and search snippets fromsearch_webas detailed in SKILL.md.\n - Boundary markers: The instructions do not specify the use of delimiters (e.g., XML tags or unique string markers) to separate untrusted data from the agent's system prompt instructions.\n
- Capability inventory: The skill utilizes web browsing, searching, and API interaction tools. It does not require or request capabilities for filesystem modification, system-level command execution, or persistent storage.\n
- Sanitization: The protocol recommends stripping boilerplate HTML content (navigation, footers, ads) to conserve tokens, but it lacks explicit filtering or sanitization to detect or neutralize malicious instructions embedded within the primary content body.
Audit Metadata