cnki-journal-toc

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill extracts and processes academic metadata from external web pages, creating a potential surface for indirect instructions to influence the agent.
  • Ingestion points: Paper titles, author names, and issue details are scraped from the CNKI journal detail and reader pages as described in SKILL.md.
  • Boundary markers: There are no explicit delimiters or boundary instructions used when the extracted data is interpolated into the final results presented to the agent.
  • Capability inventory: The skill has access to browser automation tools including mcp__chrome-devtools__evaluate_script and mcp__chrome-devtools__click to interact with the target site.
  • Sanitization: The JavaScript extraction logic performs basic trimming of strings but does not sanitize or escape content to prevent the interpretation of embedded instructions.
  • [DYNAMIC_EXECUTION]: The skill uses dynamic script assembly to interact with the browser environment.
  • The evaluate_script block in SKILL.md dynamically replaces YEAR and ISSUE placeholders with user-provided arguments at runtime.
  • While this is a common pattern for browser interaction skills, it represents the execution of dynamically generated code based on external input.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:57 AM
Security Audit — agent-trust-hub — cnki-journal-toc