promo-video

Pass

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONMETADATA_POISONINGCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes local game source code to extract logic for game-over conditions and input handling. This untrusted data is then used to generate capture scripts and browser-side patches. Ingestion points: Game source files (e.g., GameScene.js) read during adaptation. Boundary markers: Absent; no instructions to ignore embedded code directives. Capability inventory: File system access, browser automation (CDP), and shell execution (FFmpeg). Sanitization: Relies on pattern matching without validation of extracted code.
  • [DYNAMIC_EXECUTION]: The skill generates a game-specific capture script by adapting a template and uses browser-side JavaScript injection. Script Generation: Dynamically creates capture-promo.mjs based on game analysis. Browser Injection: Uses page.evaluate to override Phaser time subsystems and monkey-patch game logic.
  • [METADATA_POISONING]: The skill's author field (OpusGameLabs) contradicts the platform-provided author context (playableintelligence).
  • [COMMAND_EXECUTION]: The skill executes shell commands to verify the environment (node, ffmpeg) and performs video post-processing using a provided bash script that wraps FFmpeg.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 12, 2026, 03:52 PM