omp-publisher-connector

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically manuscript_files (DOCX/PDF) and book_metadata. It performs scanning of titles, abstracts, and body text for language detection. This represents a vulnerability surface where malicious instructions embedded in manuscripts could potentially influence agent behavior.
  • Ingestion points: manuscript_files (SKILL.md), book_metadata (SKILL.md).
  • Boundary markers: None specified for manuscript text processing.
  • Capability inventory: Subprocess execution of local Python scripts, network operations (HTTP/REST), and browser automation via Chrome DevTools MCP.
  • Sanitization: The skill performs language validation (Bahasa Indonesia check), but no explicit escaping or filtering of potentially malicious instruction sequences within the manuscript text is mentioned.
  • [COMMAND_EXECUTION]: The skill relies on executing local Python scripts (omp_headless_submitter.py and generate_standard_docx.py) via shell commands to perform monograph registration, package export, and submission tasks.
  • [DYNAMIC_EXECUTION]: The skill includes a mechanism to generate a JavaScript snippet (autofill_helper.js) designed for manual execution by the user in a browser console. Additionally, it uses the Chrome DevTools Protocol (DOM.setFileInputFiles) for visual automation and file injection during the submission process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 12:40 PM
Security Audit — agent-trust-hub — omp-publisher-connector