create-how-we-made-this-hook
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches official brand logos from Simple Icons (
cdn.simpleicons.org) and uses Google Fonts. These are well-known services commonly used for retrieving UI and video assets during development. - [COMMAND_EXECUTION]: The workflow involves local execution of FFmpeg for media processing, including trimming, scaling, and compositing various video and audio layers. It also utilizes Playwright to render HTML/CSS templates into high-quality PNG overlays for use in the video. These operations are scoped to the project's local directory and are standard for video automation tasks.
- [PROMPT_INJECTION]: The skill is designed to process external video files (
proof_film) and structured metadata (tool_stack), presenting a surface for indirect prompt injection. - Ingestion points: External media is ingested via the
source/directory, and configuration is handled through YAML inputs. - Boundary markers: The skill implements a mandatory human-approval gate (Step 3) where the VO script must be locked before any downstream generation (voiceover or lip-syncing) occurs.
- Capability inventory: Includes local file writing and video manipulation via FFmpeg and headless browser rendering via Playwright. Network access is restricted to established AI provider APIs.
- Sanitization: The skill enforces script locking in
script-locked.mdand includes steps to verify generated transcripts against the locked script to prevent processing unintended content.
Audit Metadata