DCC itch.io Skills

Two independently installable DCC-MCP Skills for safe itch.io asset acquisition and preview-gated game publishing.

Illustration generated for this repository; it contains no third-party logos, copied marketplace thumbnails, or product UI.

Skills

itchio-assets

• Browse only itch.io's official game-assets RSS feed.
• List and inspect asset projects with direct download-key ownership on the authenticated account through the official Butler daemon API.
• Download one exact owned upload without running installers or downloaded code.
• Require explicit license evidence before any download, then return a validated DCC-MCP AssetDescriptor with hashes and attribution.

This Skill does not buy assets, scrape project pages, infer a license from price or ownership, or import downloaded content into a DCC application. Version 0.1 fails closed for bundle-only ownership.

itchio-publish

• List projects owned by the authenticated creator account.
• Inspect Butler channels and their head/pending build states.
• Create a hash-bound preview receipt for a prepared game build.
• Revalidate the source tree, license provenance, Butler binary, channel parent, and preview result before an authenticated butler push.

Publishing consumes license-provenance.json produced by the marketplace game-release-package Skill. Core's asynchronous job cancellation terminates the local Butler process only; it is not a remote build rollback.

Runtime requirements

• DCC-MCP Core 0.19.91 or newer
• Python 3.7 or newer for production scripts
• itch.io Butler 15.30.0 or newer
• BUTLER_API_KEY in the local environment for authenticated operations

The API key is never accepted as a tool argument, written to a receipt, placed on a command line, or returned in tool output. The asset Skill passes it to an ephemeral, isolated Butler daemon over JSON-RPC and removes that daemon's local state after the request.

Development

python -m pip install "dcc-mcp-core>=0.19.91" "pytest>=7.4,<9" "PyYAML>=6" "ruff>=0.9" Pillow
python scripts/validate_skills.py
python -m pytest -q
ruff check .

All network and Butler unit tests use fixtures or fake processes. CI never performs a real upload or download.

License

MIT