Skip to content

PUBLICATION_CHECKLIST.md

Latest commit

 

History

History
56 lines (42 loc) · 2.46 KB

PUBLICATION_CHECKLIST.md

File metadata and controls

56 lines (42 loc) · 2.46 KB
Raw

Publication Checklist

Use this checklist before making the enhanced MCP repository public or linking it from bpy.dev.

Legal and provenance

  • Add repository license file: LICENSE.
  • Add attribution/provenance notice: NOTICE.md.
  • Confirm upstream SPDX license markers: GPL-3.0-or-later.
  • Decide final public name/package namespace (bpy-dev-mcp, bpydev-mcp, or another name).
  • Decide whether this repo remains a fork or is imported into a clean public repo.

Repo hygiene

  • Ignore local Hermes planning/session files via /.hermes/.
  • Ignore common build/cache artifacts.
  • Confirm no generated trial media or temporary outputs are tracked.
  • Confirm no obvious credentials, tokens, private keys, or private paths are tracked by the publication-prep scan. Remaining API-key references are environment-variable examples/tests, not literal secrets.
  • Confirm tracked file sizes are ordinary source/docs/fixture sizes; largest tracked files are bundled API docs and small deterministic .blend fixtures.

Product docs

  • Add security model: SECURITY.md.
  • Add developer-preview overview to readme.md.
  • Add backend documentation: readme_bpy_backend.rst.
  • Document benchmark limitations in benchmarks/token_efficiency/README.md.
  • Add final package install commands after package naming is decided.
  • Add public bpy.dev link after the website repo is live.

Functional gates

  • Install/help smoke via uv run --directory mcp blender-mcp --help.
  • Run saved-file/runtime validator smoke with Blender executable backend.
  • Run saved-file/runtime validator smoke with standalone bpy backend.
  • Run runtime API docs smoke through the unit/integration suite.
  • Run deterministic benchmark validator tests.
  • Run tracked-file secret scan.
  • Run git diff --check.

Benchmark/demo artifacts

  • Copy Trial 017 videos into the website repo or CDN/object storage.
  • Publish Trial 017 only as an exploratory case study.
  • Keep deterministic benchmark results separate from creative/VLM demo results.
  • Do not claim definitive superiority from a single creative trial.

Release notes draft

Developer preview focus:

  • saved-file/headless Blender MCP execution;
  • standalone bpy backend option;
  • runtime Blender Python API lookup;
  • subprocess/capture/transport hardening;
  • deterministic benchmark tasks and telemetry scaffolding;
  • exploratory creative comparison tooling.