[lwaldron]

Closes #37.

Adds skills/validate-protocol: run scripts/validate_protocol.py, classify every error it can emit as either mechanical (fix it and say what changed) or a judgement call (state the choice and stop), re-run until only judgement calls remain, and report. The classification tables cover every message in the validator and the models, with the wording checked against the invalid fixtures and pydantic output.

Two restraints run through it: never invent a DOI, author, date, review, or status, and never edit Materials, Steps, Changes bullets, or review Notes text. Several errors that look mechanical are not, and the tables say when: a legacy citation field, a frontmatter version ahead of the top History entry, and a *No reviews yet.* placeholder when a review exists somewhere.

The issue named the R validator; the skill targets the Python one from ADR 0015, and covers the tests and materials-conditions rules from ADR 0016 and 0017.

Sibling PRs for #35 and #36 each add a bullet to skills/README.md at the same spot, so the second and third to merge will need a trivial conflict resolution there.

🤖 Drafted with Claude Code

[claude]

⚠️ Code review skipped — your organization has no extra usage available to pay for this review.

If your organization's extra usage balance is empty, an organization admin can add extra usage credits at claude.ai/admin-settings/usage. If its monthly spend limit was reached, an admin can raise it on the same page. If neither applies, contact Anthropic support.

Once extra usage is available, someone with write access to this repository can comment @claude review on this pull request to trigger a review.