chore(release): prepare v5.6.1

[JeremyDev87]

Summary

Bump the workspace version from 5.6.0 → 5.6.1 and publish the CHANGELOG entry for the HUD hotfix cycle.

v5.6.1 is a hotfix closing the remaining gaps in the v5.6.0 HUD Statusbar Wave cycle:

• Wave 1-B heal timestamp leak (fix(hud): stop heal_stale_state leaking stale sessionStartTimestamp (Wave 1-B) #1487, merged as cd2c9d3) — heal_stale_state preserved sessionStartTimestamp for "audit / forensics", but resolve_duration read the same field as a fallback when stdin lacked total_duration_ms. A manual-fix marker + stale timestamp therefore rendered enormous durations like 322h52m for brand-new sessions. The timestamp is now relocated into _healedFromSessionStartTimestamp.

• Wave 3b — complete Wave 3 integration (feat(hud): complete Wave 3 integration for 1-D/2-A/2-D/2-E (Wave 3b) #1488, merged as 64edfac) — commit bd78195 wired Wave 2-B velocity and 2-C cache savings into format_status_line but left four sibling Wave modules as dead code: hud_buddy, hud_rainbow, hud_context_bar, and hud_layout. Their unit tests passed, but format_status_line never called them. v5.6.1 hoists all four as top-level imports and refactors format_status_line to build (name, priority, text) segments consumed by fit_segments, finally delivering Wave 2-A breathing face, Wave 2-D rainbow coloring (opt-in via CODINGBUDDY_HUD_RAINBOW=1), Wave 2-E smart context bar, and Wave 1-D adaptive layout — all features v5.6.0 advertised but shipped inert.

Bump surface (10 files)

File	Change
apps/mcp-server/package.json	5.6.0 → 5.6.1
apps/mcp-server/src/shared/version.ts	VERSION const
packages/rules/package.json	5.6.0 → 5.6.1
packages/claude-code-plugin/package.json	version + peerDependencies.codingbuddy
packages/claude-code-plugin/.claude-plugin/plugin.json	plugin schema version
packages/claude-code-plugin/README.md	generated version stamp
packages/claude-code-plugin/namespace-manifest.json	yarn build output
.claude-plugin/marketplace.json	plugins[0].version
yarn.lock	peerDep resolution
CHANGELOG.md	new [5.6.1] - 2026-04-11 section

Test plan

• ./scripts/bump-version.sh 5.6.1 executed, all 7 version files confirmed ✅ (the yarn.lock clean false-positive is just the script's post-install diff check — the actual delta is codingbuddy: ^5.6.0 → ^5.6.1 in peerDependencies, expected)
• yarn workspace codingbuddy-claude-plugin build — namespace-manifest.json + README.md refreshed
• python3 -m pytest packages/claude-code-plugin/tests/test_hud.py tests/test_hud_session.py — 153 passed on post-merge master
• CI: e2e-plugin-hooks, install-dependencies, e2e-plugin-docker, Vercel preview
• User review + merge
• User executes release: git tag v5.6.1 && git push origin v5.6.1
• User runs GitHub Release / publish workflow

Follow-on (not in this PR)

• Tag + GitHub Release creation — user-driven per CLAUDE.md release policy
• Post-release validation: verify v5.6.1 HUD behavior in a fresh session (buddy face changes with phase, CODINGBUDDY_HUD_RAINBOW=1 shows gradient, context bar renders)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
codingbuddy-landing	Ready	Preview, Comment	Apr 11, 2026 4:07pm

[JeremyDev87]

Code review — PR #1489

Verdict: LGTM with one minor CHANGELOG accuracy nit. Release prep is mechanically clean — every version-bearing file is consistently bumped, the regenerated namespace-manifest.json matches the unchanged command surface, the yarn.lock peerDep resolution is correct, and the CHANGELOG correctly attributes the two underlying fixes to the right merged commits. All 25 CI checks are green. The single finding below is a test-count number that is overstated relative to the actual diff in #1488 — non-blocking, but worth correcting if you'd like the CHANGELOG to be a faithful record.

Findings

Critical

None.

High

None.

Medium

None.

Low

• CHANGELOG.md overstates the Wave 3b new-test count. The new ### Test Coverage section claims "17 new Wave 3b integration tests across four classes (TestWave2ABreathingFaceIntegration, TestWave2EContextBarIntegration, TestWave2DRainbowIntegration, TestWave1DAdaptiveLayoutIntegration)". The actual git show 64edfac diff adds 14 new test methods across those four classes (4 + 3 + 4 + 3), which matches the commit message of #1488 verbatim ("14 new tests split across four classes"). The Wave 1-B count ("5 new") is correct (3 unit + 2 integration). Suggested fix in a follow-up: change 17 → 14 in CHANGELOG.md. This is cosmetic and does not affect runtime behavior, so it should not block tagging — but if you're amending the release-prep commit before tagging, it's a one-character fix.

Nit

• CHANGELOG date stamp is [5.6.1] - 2026-04-11. The two underlying commits (cd2c9d3, 64edfac) are dated Apr 12 00:26 +0900 and Apr 12 00:35 +0900, which is Apr 11 15:26 / 15:35 UTC — so 2026-04-11 is internally consistent in UTC. Per Keep-a-Changelog convention the date should reflect the actual release-tag day; if you tag on 2026-04-12 (today, KST), you may want to bump the CHANGELOG date to match. Pure judgment call — leave as-is if you prefer the "PR-prepared on" semantic.

Verification performed

• gh pr view 1489 --json title,body,labels,headRefName,baseRefName,state,url,headRefOid,additions,deletions,changedFiles — confirmed PR is OPEN against master from chore/release-v5.6.1 at 0a2bec5, +33 / -10, 10 changed files.
• gh pr diff 1489 — read the entire diff. All version bumps are 5.6.0 → 5.6.1 with no collateral edits.
• gh pr view 1487 --json … and gh pr view 1488 --json … — confirmed both PRs are MERGED and that the merge-commit OIDs (cd2c9d3226c4372caaa8692be44aba7298023236, 64edfac6ba2286e58e2871cf290590b5ff2ef6e5) match what the PR body and CHANGELOG cite. The CHANGELOG hyperlinks point to the right PR numbers under JeremyDev87/codingbuddy.
• git log cd2c9d3 -1 --stat and git log 64edfac -1 --stat — read both commit messages and per-file stats. Compared the prose against the new [5.6.1] section in CHANGELOG.md and found the only material discrepancy is the 17 vs 14 Wave-3b new-test count called out above.
• git show 64edfac -- packages/claude-code-plugin/tests/test_hud.py — counted the + def test_* and +class Test* lines actually added to confirm 4 + 3 + 4 + 3 = 14 new Wave 3b integration tests, matching the commit message and contradicting the CHANGELOG's 17.
• git show cd2c9d3 -- packages/claude-code-plugin/tests/test_hud.py packages/claude-code-plugin/tests/test_hud_session.py — counted Wave 1-B additions: TestHealedStateDurationDoesNotLeak (2 integration) + 3 unit functions in test_hud_session.py = 5 total, which matches CHANGELOG's "5 new Wave 1-B regression tests".
• Read packages/claude-code-plugin/namespace-manifest.json — verified the regenerated manifest only changed generatedAt (2026-04-11T14:34:02.760Z → 2026-04-11T16:03:33.849Z); the command list is unchanged, which is correct because v5.6.1 ships no new slash commands.
• Grep '5\.6\.0' across the repo — confirmed the only remaining 5.6.0 strings outside .claude/worktrees/ (untracked working copies) and 5.6.0 substrings of unrelated deps (chalk@npm:^5.6.0, actions/setup-python … # v5.6.0) are: (a) the five README*.md "What's New in v5.6.0 — HUD Statusbar Wave" headers, which is consistent with the team convention of not updating README translations on hotfix releases, and (b) packages/claude-code-plugin/tests/test_hud.py:325, which is a docstring inside the heal-fix test that literally reproduces the buggy v5.6.0 status line — that's intentional as a regression-test fixture. No stale runtime references.
• gh pr checks 1489 — all 25 checks pass: build-check, circular-check, install-dependencies, landing-* (7 checks), lint-check, plugin-* (8 checks: build/circular/format/lint/security/test/type/validate-commands), prettier-check, rules-validation, security-check, type-check, unit-test-check, Vercel, Vercel Preview Comments. Zero red, zero pending.
• git log --oneline master..chore/release-v5.6.1 — confirmed the branch contains exactly one commit (0a2bec5 chore(release): prepare v5.6.1), no stray work.

Notes

• Did not check out the branch, run tests, or attempt any modifications (review-only role).
• The PR body's Test plan checkbox python3 -m pytest … — 153 passed is the count for the two test files explicitly listed (test_hud.py + test_hud_session.py), not the full plugin suite — that's consistent with running just those two files.
• Once the test-count nit is decided (amend or punt), this PR is ready to merge. After merge, per CLAUDE.md release policy, tag creation (git tag v5.6.1 && git push origin v5.6.1) and the GitHub Release publish are user-driven — do not let an agent execute them.