The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external binaries for media processing and testing.
tools/beatmap.py uses subprocess.run to call ffmpeg and ffprobe for music analysis.
tools/encode.sh invokes ffmpeg for final film assembly.
engine/tests/extra.mjs and engine/tests/kit_check.mjs use execFileSync to run internal project CLI tools.
[EXTERNAL_DOWNLOADS]: External dependencies are fetched from well-known sources.
setup.sh uses npm ci to install pinned packages from the official npm registry.
tools/rp_pod/setup.sh fetches the Node.js runtime from nodejs.org and verifies it using SHA256 checksums.
[DYNAMIC_EXECUTION]: Runtime code execution is used for path tracing and remote rendering workflows.
engine/src/host.mjs utilizes puppeteer-core to drive headless Chrome for executing custom WGSL compute shaders on the GPU.
tools/rp provides a mechanism to synchronize project code and execute rendering tasks on remote Runpod GPU instances via SSH and rsync.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data as part of its rendering pipeline.
Ingestion points: tools/beatmap.py (audio tracks) and engine/src/scene.mjs (scene JSON, PTM meshes, and PTV volume data).
Boundary markers: Absent; data is passed directly to specialized processing tools.
Capability inventory: subprocess.run (ffmpeg) in tools/beatmap.py, browser-based shader execution in engine/src/host.mjs, and remote shell access in tools/rp.
Sanitization: Absent; the skill relies on the robustness of the underlying tools (ffmpeg, Puppeteer) and its custom binary parsers to handle input safely.

code-based-films