The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/scaffold.py and references/build_hf_clips.py scripts utilize subprocess.run() to invoke npx hyperframes init, which initializes the project environment and installs necessary dependencies.
[REMOTE_CODE_EXECUTION]: The skill's HTML templates incorporate a script tag to load the GSAP animation library from the well-known JSDelivr CDN (https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js).
[DYNAMIC_EXECUTION]: The skill dynamically generates executable HTML and JavaScript by interpolating data from JSON specifications, SRT subtitle files, and user-defined fragment files into predefined templates. Specifically, scripts/scaffold.py inlines raw code fragments into the final composition file.
[DATA_EXPOSURE]: Documentation within SKILL.md identifies ~/.codex/auth.json as a sensitive file containing credentials required by the make_cover.py script for authentication.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (JSON, SRT) that is interpolated into the generated output, creating a potential surface for injection.
Ingestion points: Configuration data is read from segments.json, spec.json, and SRT subtitle files.
Boundary markers: The templates lack explicit delimiters or warnings to prevent the agent from following instructions embedded in the processed data.
Capability inventory: The skill possesses capabilities for file system operations and command execution via shell subprocesses.
Sanitization: While some basic HTML escaping is applied in scripts/scaffold.py, it is not consistently implemented across all scripts and data fields.

wjs-overlaying-video