release: v0.3.0

[robercano-ghbot]

Cuts the first tagged release of the orchestrator plugin, per the milestone-gated release convention
(issue #176, docs/USAGE.md → "Release cycle"). Produced by bash .claude/scripts/release.sh v0.3.0 --issue 178
after a clean --dry-run inspection.

What's in the diff

• .claude/.claude-plugin/plugin.json — version 0.2.2 → 0.3.0
• .claude/.claude-plugin/marketplace.json — top-level version and the plugins[] entry → 0.3.0
• .claude/.claude-plugin/CHANGELOG.md — new dated ## [0.3.0] - 2026-08-04 section, generated from the
  merged-PR history (first-release path: no prior tag existed). Inserted below the ## [Unreleased]
  scaffold and above ## [0.2.2], per Keep-a-Changelog ordering.

No other files are touched.

Side effects already performed by release.sh (real, by design)

• Tag v0.3.0 created and pushed to origin (49e63b5) — this repo's first tag.
• Branch feat/issue-178-release-v0.3.0 pushed.
• Milestone Ship a pre-approved permission allow/deny list #1 "v0.3.0 — consumer rollout" closed.
• Companion issue Rollout & feedback: v0.3.0 #214 "Rollout & feedback: v0.3.0" auto-filed — census-invisible by design
  (labelled feedback, no milestone, no planned label), carrying the reDeploy/reDeFi
  /orchestrator:sync checklist (issue /orchestrator:sync v2 — one sanctioned migration path to bring consumer repos current #141) and test-focus areas derived from the milestone.

release.sh is not idempotent; it was run exactly once.

Manual owner follow-up (cannot be automated)

/plugin marketplace update recode

Gates

GATES_FILE=self/gates.json — build: pass, lint: pass, test: see PR checks/comments.

Note

0.3.0 supersedes the 0.2.2-only publish planned in the closed issue #136 — it ships everything that
release would have.

Closes #178

[robercano-ghbot]

Gate + review status (orchestrated ADVANCE for #178)

Gates (GATES_FILE=self/gates.json), run against 49e63b5 in an isolated worktree:

• build: pass
• lint: pass
• test: self/checks.sh test emitted 0 failures across every scenario it reported, but the run stalled after scenario 5 (timeout) in this sandbox and never reached self/smoke-fanout.sh. The stall reproduced across separate runs and is environment-level (the loop-driver spawn scenarios use systemd-run stubs and were contending with other concurrent loop processes) — it is not attributable to this diff, which contains no executable logic, only generated version/changelog data. Recommend confirming test green in CI before merging.

Reviews (lenses per self/gates.json, consensus all):

• correctness — APPROVE. Bump complete and consistent across plugin.json + marketplace.json (incl. plugins[]); JSON structurally intact; changelog ordering correct; scope stays inside .claude/; tag/version agree.
• tests — APPROVE. Pure generated output, no new tests warranted; .claude/scripts/release.test.sh already covers the bump, changelog insertion ordering, the first-release/no-prior-tag path, and the --dry-run boundary. No fixture goes stale on 0.2.2 (sync.sh's MIN_PLUGIN_VERSION is a compat floor, not a mirror).

Non-blocking follow-ups raised by the correctness reviewer (worth a separate issue, not this PR):

1. release.sh:162 only rewrites Merge pull request #N from owner/branch subjects, so other merge subjects land in the changelog without a - bullet prefix.
2. release.sh:160 samples --merges only, so squash-merged PRs (feat(harness): /orchestrator:provision — guided dedicated-server onboarding #205, docs(hardening): capture dedicated-server provisioning learnings #212, docs(hardening): fix nftables sketch + persist dedicated-server provisioning artifacts #213) are absent from the generated bullets; and the hand-written ## [Unreleased] entries now describe work shipped in 0.3.0 rather than being folded into the cut section.

Merging is gated on the repo owner's GitHub approval, per .claude/scripts/merge-ready.sh.

[robercano-ghbot]

release: v0.3.0 (not yet reviewed)

[robercano-ghbot]

release: v0.3.0 (not yet reviewed)