skills/notque/vexjoy-agent/publish/Gen Agent Trust Hub

publish

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute standard Unix utilities (grep, find, ls) and custom Python scripts for publishing tasks. The batch-edit.md reference specifically allows for bulk find-and-replace operations across the local filesystem using regular expressions. To mitigate risk, the skill requires a clean Git status before applying changes to ensure a rollback path exists.
  • [INDIRECT_PROMPT_INJECTION]: The link-audit.md workflow and the link_scanner.py script read and parse markdown content from the content/ directory. This creates a surface where malicious instructions or links embedded in post data could influence the agent's behavior during analysis.
  • Ingestion points: Markdown files in the content/ directory are read by link_scanner.py, wordpress-upload.py, and pre-publish.md.
  • Boundary markers: The skill correctly identifies TOML (+++) and YAML (---) frontmatter fences as structural boundaries.
  • Capability inventory: The skill has Read, Write, Edit, and Bash tools, enabling filesystem modifications and script execution. It also performs network requests to WordPress and Google APIs.
  • Sanitization: The scripts use standard parsing for frontmatter and basic regex for markdown-to-HTML conversion, but do not implement advanced sanitization of the content before processing.
  • [DATA_EXFILTRATION]: The link_scanner.py script performs network requests (HEAD and GET) to arbitrary external URLs found in the blog content to verify their health. While this is a core feature of a link auditor, it allows for requests to non-whitelisted domains. Furthermore, the WordPress scripts (wordpress-upload.py, wordpress-edit-post.py, and wordpress-media-upload.py) read sensitive Application Passwords from a local ~/.env file and transmit them via Basic Authentication to the user-defined WORDPRESS_SITE.
  • [EXTERNAL_DOWNLOADS]: The skill requires external Python libraries (requests and markdown) to be installed in the environment for its scripts to function. The documentation provides instructions for installing these dependencies via pip.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 03:46 PM