word-processor-expert

Warn

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The provided 'Find & Replace' implementation uses new RegExp(searchTerm, flags) with searchTerm derived directly from user-controlled input. This is vulnerable to Regular Expression Denial of Service (ReDoS) attacks, where a malicious pattern can cause the application or browser to freeze due to catastrophic backtracking.\n
  • Evidence: Found in the FindReplace component logic in skill.md.\n- [REMOTE_CODE_EXECUTION]: The PDF export API route utilizes Puppeteer's page.setContent(html) to render documents. Because the HTML is passed directly from the client without sanitization, it is vulnerable to Cross-Site Scripting (XSS). An attacker could inject malicious scripts into a document that execute within the server-side browser context during export, potentially accessing local files or making unauthorized network requests.\n
  • Evidence: Located in the Puppeteer POST route implementation in skill.md.\n- [PROMPT_INJECTION]: The skill creates a system for ingesting and processing untrusted document text. When this content is extracted and passed to subsequent agent steps or processing functions without proper boundary markers, it acts as a vector for indirect prompt injection.\n
  • Ingestion points: Document content retrieved via editor.getText(), editor.getJSON(), and editor.getHTML() in skill.md.\n
  • Boundary markers: Absent. The skill does not provide instructions for the agent to distinguish between system commands and user-provided document text.\n
  • Capability inventory: Includes the ability to modify document state, perform text replacement, and trigger external file exports via PDF and DOCX libraries.\n
  • Sanitization: Not explicitly present in the provided code snippets for text processing or rendering.\n- [EXTERNAL_DOWNLOADS]: The skill specifies dependency versions for jspdf (v3.0.3) and docx (v9.5.1) that do not currently exist in the official NPM registry. Referencing non-existent versions of libraries can be a risk factor for supply chain attacks if those versions are later published by unauthorized actors.\n
  • Evidence: Mentioned in the 'Current Stack' section of skill.md.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 13, 2026, 06:36 AM
Security Audit — agent-trust-hub — word-processor-expert