The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary text provided by the user (articles, notes, etc.) to create video content. This input is stored and subsequently used to generate storyboards and instructions for sub-agents (workers). Maliciously crafted input could potentially attempt to influence the behavior of these sub-agents during the frame-building process.
Ingestion points: User input is saved verbatim to capture/extracted/visible-text.txt in Step 1.
Boundary markers: The skill does not explicitly mention the use of delimiters or 'ignore' instructions when interpolating this user content into prompts for sub-agents.
Capability inventory: The agent and its sub-agents have the capability to execute shell commands (via npx), read and write local files, and generate dynamic HTML/JS content.
Sanitization: No explicit sanitization or filtering of user-provided text is described before it is used in the generation pipeline.
[COMMAND_EXECUTION]: The skill and its associated scripts make extensive use of the npx hyperframes CLI tool and system utilities like ffmpeg and ffprobe via spawnSync. This execution is required for the intended purpose of video project management, asset processing, and final rendering.
Evidence: Multiple steps in SKILL.md and scripts such as assemble-index.mjs invoke these tools to handle media files and project initialization.
[EXTERNAL_DOWNLOADS]: The generated HTML compositions for video frames and captions include references to the GSAP library hosted on a well-known public CDN (jsdelivr.net).
Evidence: scripts/assemble-index.mjs and scripts/captions.mjs contain script tags pointing to https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js, including SRI integrity hashes to ensure the library has not been tampered with.

hyperframes-faceless-explainer