The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes numerous external input files that serve as instructions for the animation production process.
Ingestion points: Processes various JSON and Markdown files provided in the 09-导演/ (director) and 10-素材/ (assets) directories, including animation-brief.json, renderer-brief.md, and edl.json.
Boundary markers: The instructions mandate strict adherence to an AST (Abstract Syntax Tree) allow-list and require explicit validation of inputs via SHA-256 hash checks.
Capability inventory: Utilizes local Node.js scripts for file system management, project scaffolding, rendering, and QA evidence generation.
Sanitization: Includes multiple verification steps, such as verify-renderer-handoff.mjs and verify-renderer-workspace.mjs, alongside a mandatory human approval step (viewer_verdict).
[EXTERNAL_DOWNLOADS]: The reference HTML templates for animation components fetch a legitimate third-party library.
Details: The GSAP animation library is loaded via script tags targeting https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js (a well-known, trusted CDN service).
[COMMAND_EXECUTION]: The workflow relies extensively on the execution of local utility scripts located within the skill's project structure.
Evidence: Multiple commands are described, such as node scripts/prepare-renderer-handoff.mjs, node scripts/render-episode-<renderer>.mjs, and node scripts/seal-renderer-workspace.mjs.
[DYNAMIC_EXECUTION]: The skill automates the creation and modification of animation project workspaces based on input specifications.
Evidence: The workflow involves creating unsealed scaffolds (create-renderer-workspace.mjs) and implementing approved extensions within the workspace before sealing it for production.

laohan-donghua