clipforge-video

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, creating a potential attack surface for indirect prompt injection.
  • Ingestion points: Data enters the context from product pages (via URL), stock metadata, transcripts, and image contact sheets (SKILL.md).
  • Boundary markers: The skill includes a specific "Hard Rule" (Rule 7) explicitly instructing the agent to treat fetched content as data rather than instructions and to disregard any embedded bypass attempts like "ignore previous instructions."
  • Capability inventory: The skill has access to local file execution via CLI tools and network operations via MCP tools and HTTP APIs.
  • Sanitization: The skill relies on behavioral instructions for the agent to describe rather than obey untrusted input, rather than implementing programmatic filtering.
  • [CREDENTIALS_UNSAFE]: The skill correctly manages sensitive information by instructing the use of environment variables (e.g., CLIPFORGE_LLM_API_KEY) instead of hardcoding secrets or suggesting insecure storage.
  • [COMMAND_EXECUTION]: The skill utilizes local CLI commands and Node.js scripts for video processing. These are static paths to local binaries provided by the vendor, representing standard functionality for a developer-oriented tool.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 07:42 AM
Security Audit — agent-trust-hub — clipforge-video