personal-search

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill relies exclusively on platform-integrated MCP tools (yuque_search and yuque_get_doc) to perform its functions, which is consistent with its stated purpose of personal knowledge base searching.
  • [SAFE]: There are no external network calls, hardcoded credentials, or command-line executions involving sensitive system areas.
  • [SAFE]: The instructions focus on user-centric tasks such as keyword extraction and document synthesis, with no attempts to bypass safety filters or override agent behavior.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes content from external Yuque documents.
  • Ingestion points: Document content retrieved from Yuque via the yuque_get_doc tool (SKILL.md).
  • Boundary markers: Absent; there are no specific delimiters to separate untrusted document content from the agent's internal instructions.
  • Capability inventory: Reading document content and summarizing it for output to the user.
  • Sanitization: No explicit validation or filtering of document content is mentioned in the workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 06:41 AM