clone-video-ad

Pass

Audited by Gen Agent Trust Hub on Aug 13, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external video files and their transcripts to generate advertising prompts, creating a surface for indirect prompt injection. Ingestion points: source video files (SKILL.md Step 1) and API transcripts (SKILL.md Step 2). Boundary markers: Absent; no delimiters or instructions to ignore embedded commands are used. Capability inventory: Shell execution of ffmpeg, network operations via curl, and file system access. Sanitization: Absent; no validation or escaping of transcript content is described before interpolation into prompts.
  • [EXTERNAL_DOWNLOADS]: The skill installs ffmpeg via Homebrew and openai-whisper via pip. It also provides instructions for fetching scripts and documentation from the vendor's repository (github.com/novoads/claude-code-ads). These sources are well-known or vendor-controlled.
  • [COMMAND_EXECUTION]: Executes ffmpeg for media processing and curl for interacting with the Novoads API, which are essential and legitimate operations for the skill's declared purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 13, 2026, 08:17 PM
Security Audit — agent-trust-hub — clone-video-ad