tuzi-short-video

Warn

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill attempts to locate a sensitive API key (TUZI_API_KEY) by searching through environment files in both the local project and the user's home directory (.tuzi-skills/.env and $HOME/.tuzi-skills/.env).
  • [COMMAND_EXECUTION]: The skill executes external scripts using npx -y bun. The target script path (main.ts) is determined dynamically by resolving the location of a dependent skill (tuzi-video-gen), which constitutes dynamic loading from a computed path.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user content (text, articles, or descriptions) and transforms it into visual prompts for a video generation model without sanitization or protective boundary markers.
  • Ingestion points: User input analysis in SKILL.md (Step 1).
  • Boundary markers: None identified in the prompt generation logic.
  • Capability inventory: Shell command execution via npx in SKILL.md (Step 4).
  • Sanitization: No evidence of escaping or filtering for the generated video prompts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 7, 2026, 06:51 AM