longform-publish-folder

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation provides command-line examples for users to manually trigger the Node.js script or schedule it via cron.
  • [EXTERNAL_DOWNLOADS]: The script communicates with the Late API (zernio.com) for media uploads and publishing, and the Google Gemini API for metadata generation.
  • [DATA_EXFILTRATION]: The script reads the project's 'SOCIALS.md' file to retrieve the Profile ID and Late Account ID necessary for authentication with the publishing service.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as video filenames and user-provided topic strings are interpolated into the prompt used for AI metadata generation.
  • Ingestion points: Video filenames in the watch folder and the '--topic' argument in 'scripts/longform-publish-folder.js'.
  • Boundary markers: Absent. Input data is placed directly into the natural language prompt.
  • Capability inventory: File system access for reading videos and writing sidecar status files, along with network access for uploading content.
  • Sanitization: Filenames undergo basic character replacement of underscores and hyphens, but no formal sanitization or escaping is applied to prevent instruction overrides in the prompt context.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 08:20 AM
Security Audit — agent-trust-hub — longform-publish-folder