klap-generate-shorts

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The desktop notification system uses string interpolation for building shell commands.
  • Evidence: The send_desktop_notification function in scripts/klap_generate_shorts.py uses f-strings to insert the message and title variables directly into quotes for osascript and powershell commands.
  • Risk: This can lead to command injection if the input strings contain malicious characters that break out of the quotes.
  • [PROMPT_INJECTION]: The skill processes external data from both user input and API responses, creating an indirect prompt injection surface.
  • Ingestion points: Video URLs from the command line and video metadata (e.g., titles, captions) from the Klap API.
  • Boundary markers: No explicit markers are used to separate untrusted external data from the skill's operational logic.
  • Capability inventory: The skill is capable of executing shell commands (subprocess.run), writing to a database (psycopg2), writing to the local file system (open), and making network requests (requests).
  • Sanitization: Database interactions are properly sanitized with parameters, but shell-based notifications lack sanitization for special characters.
  • [EXTERNAL_DOWNLOADS]: Downloads video files from Klap's API infrastructure.
  • Evidence: Uses the requests library to fetch MP4 files from URLs provided by the API (often hosted on Google Cloud Storage).
  • [SAFE]: Secrets and credentials are managed through environment variables and .env files, which is a standard security practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 08:20 AM
Security Audit — agent-trust-hub — klap-generate-shorts