Pin the toolchain to kaizen-loop v0.1.4 and verifier v0.1.1

[s-hiraoku]

Pins the onboarding toolchain to a newly verified compatible set.

component	was	now	commit
kaizen-loop	v0.1.3	v0.1.4	5e03791
builder-agent	v0.1.0	v0.1.0	unchanged
verifier	v0.1.0	v0.1.1	28fc7f4

What each tag carries

kaizen-loop v0.1.4 derives the verifier expectedRef from the pinned
toolchain manifest (kaizen-loop#362). Before this, a pinned install failed the
verifier freshness preflight as soon as verifier main moved ahead of the
pinned tag — the config said refs/heads/main while the installed build was a
tag. The config schema now accepts refs/tags/ as well as refs/heads/, and
annotated tags are peeled with ls-remote ... ^{}.

verifier v0.1.1 restores the mechanical verification evidence the verifier
is supposed to weigh (verifier#221, fixes kaizen-loop#357). verify_commands
is populated from the verification log block and evidence_grade is upgraded
to executed when those commands are present.

Verification

kaizen-loop at 5e03791 — 672 tests passed, typecheck clean,
check:dist current.

verifier at 28fc7f4 — the repository's full CI contract from a clean
checkout: typecheck, test, build, test:package-entry,
test:built-cli, schema:check, eval, fixture-run, eval:relaxations,
the semantic eval CI gate, and git diff --exit-code clean.

Producer/parser pairing. The verifier's parser only extracts commands from
a <verification_logs_data> block wrapping a ```markdown fence, and
returns empty silently on any mismatch — which would look exactly like the
original bug. kaizen-loop's real producer output was fed to the real parser:
block found, fence matched (length 4), records parsed. The two halves pair.

Clean install from this manifest into an isolated prefix, all three
components resolving into $KAIZEN_HOME:

kaizen-loop    v0.1.4 (5e03791)
builder-agent  v0.1.0 (98fc019)
verifier       v0.1.1 (28fc7f4)

kaizen doctor --repair --project <registered slug> — 12/12 checks pass,
including gh auth and verifier agent, with verifier main (28fc7f4)
deliberately ahead of the pinned tag. That is the exact condition that failed
before kaizen-loop#362.

End-to-end run on a Rust repository (s-hiraoku/topcoat-sandbox), which is
the evidence for kaizen-loop#357:

"evidence_grade": "executed",
"verify_commands": [
  {"command": "cargo test",   "exit_code": 0},
  {"command": "cargo clippy", "exit_code": 0}
],
"verdict": "open_pr", "final_verdict": "mergeable", "risk": "low"

verify.log is 338 lines with test result: ok. 3 passed. Before v0.1.1 this
same payload had verify_commands: [] and evidence_grade: "reported".

Known limitation, stated deliberately

Configs generated by kaizen init under v0.1.4 use expectedRef: refs/tags/..., which does not parse under v0.1.3 or earlier — the older
schema regex only accepts refs/heads/. Adopters cannot downgrade after
running init with this set. This is recorded in the v0.1.4 tag message.

Not covered by this PR

The smoke run above reached a verifier verdict but could not publish a PR:
the publication broker refuses the push with a generic failure. Filed
separately as #217 (workspace mode) and the broker-side temp-ownership
problem; kaizen-loop#368 records the same failure observed from inside a run.
Those are broker/workspace defects, independent of the three pinned tags.

Summary by CodeRabbit

• Updates
  • Updated onboarding content versions for the Kaizen Loop and Verifier components.
  • Kept the Builder Agent version unchanged.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: kaizen-agents-org/coderabbit/.coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: a1231a09-77df-4194-8de9-ac9204032862

📥 Commits

Reviewing files that changed from the base of the PR and between 376afb0 and d969e07.

📒 Files selected for processing (1)

• onboarding/versions.json

---

📝 Walkthrough

Walkthrough

The onboarding manifest updates kaizen-loop from v0.1.3 to v0.1.4 and verifier from v0.1.0 to v0.1.1. The builder-agent version remains v0.1.0.

Changes

Onboarding version pins

Layer / File(s)	Summary
Update onboarding component pins onboarding/versions.json	The manifest advances the kaizen-loop and verifier versions. The builder-agent version is unchanged.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related issues

• Advance pinned onboarding set to kaizen-loop release with verifier tag pinning #211 — Requests the same kaizen-loop and verifier version updates in onboarding/versions.json.
• [scout] Align verifier CLI version with release tag verifier#223 — Addresses verifier release/version metadata aligned with the updated onboarding pin.

Possibly related PRs

• kaizen-agents-org/verifier#36 — Provides verifier changes associated with the updated verifier version.
• kaizen-agents-org/.github#118 — Updates the same onboarding version manifest.
• kaizen-agents-org/.github#204 — Updates the kaizen-loop version in the same manifest.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely identifies the two toolchain version pins updated by this pull request.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch release-v0-1-4

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[s-hiraoku]

PR Guardian

Merge-ready. No fixes were needed — no actionable feedback was raised.

Feedback inventory

source	result
Verify organization contracts (required)	✅ SUCCESS (57s)
CodeRabbit	✅ "No actionable comments were generated in the recent review."
Codex	⚠️ did not review — account usage limit reached
Inline review threads	0 total, 0 unresolved (incl. outdated)
Human reviews	none requested

Final gate

Two snapshots taken 35s apart after the stabilization wait were identical:

head        d969e07 (unchanged across both)
isDraft     false
mergeable   MERGEABLE
mergeState  CLEAN
checks      CodeRabbit=SUCCESS, Verify organization contracts=SUCCESS
unresolved  0

Base is current with main (376afb0), which matters here because the branch
protection sets strict: true. Required conversation resolution is enabled and
satisfied — there are no threads to resolve.

Notes on two items I did not treat as blockers

Codex did not review. It is not in the required status checks
(required_status_checks.contexts contains only Verify organization contracts), and its absence is an external account limit rather than a pending
verdict on this diff. Recording it explicitly rather than reporting a clean
bot sweep.

CodeRabbit's "no linked issues found". The two issues this release closes,
kaizen-loop#362 and kaizen-loop#357, are both already CLOSED — they were
fixed in the component repositories and closed by their own PRs. This PR only
bumps the pinned manifest, so adding a closing keyword would be inaccurate. No
change made.