Publish gate → default branch + docs/skills alignment + 0.2.1 baseline

[thorwhalen]

Pre-fleet-migration release. Three things:

1. Publish gates on the repo's actual default branch

Both uv-ci.yml and the inline github_ci_uv.yml switched the publish if: from a literal master || main to:

github.ref == format('refs/heads/{0}', github.event.repository.default_branch)

This matches the github-pages job and is correct for repos whose default branch is neither master nor main (which previously would never publish), and avoids publishing from a stray non-default master/main branch. Combined with the prior fix, publish now requires: default branch + validation green + not [skip ci] + publishing enabled. test_workflow_gates.py / test_uv_workflow_template.py updated to pin it.

2. Version baseline → 0.2.0 (publishes 0.2.1)

The CI auto-bump increments on publish, so setting the baseline to 0.2.0 makes the next publish ship 0.2.1, as requested.

3. Docs & skills aligned with the new wads way

For the upcoming migration of 10–20 repos, the guidance must be correct:

• wads-migrate SKILL.md (the migration guide; symlinked into ~/.claude/skills/): removed the stale secrets: inherit claim and the old "curated env-var list" text; documented wads-secrets add and the transport-vs-env-policy two-layer model + new Secrets section.
• README.md: new "Configure Secrets" section (wads-secrets, [tool.wads.ci.env], the superset) and a note that publishing happens on the default branch only when validation passes.
• CLAUDE.md: replaced legacy github_ci_publish_2025.yml references with the stub/uv-ci.yml; documented the export-ci-env step and the validation-gated, default-branch publish flow.

wads-ci-fix and wads-ci-sweep skills were reviewed and are already accurate.

Full suite: 286 passed.

As with the earlier reusable-workflow fixes, this PR's own CI runs against the current @master workflow; the publish/branch-gate change takes effect for consumers once merged. The post-merge run on master is the real verification and will publish 0.2.1.