Rename murat → qa-lead and make it @-addressable

[don-petry]

Summary

Companion to petry-projects/.github#752 (the standard + schema). Applies the new "a persona is a role, not a person" rule (§1.6) and the new addressing contract (§4.1) to the worked example.

Renames via git mv (history preserved):

personas/murat/ → personas/qa-lead/
evals/murat/    → evals/qa-lead/

plus id / name / opt_out_label / canary.agent / evals.path and the eval case ids.

The role, not the cast list

The vendored agent this persona wraps has a person-name upstream. It keeps it; we do not import it. @petry-projects/qa-lead says who is being addressed and why in a PR comment — the upstream cast list does not — and swapping the agent underneath now never changes how the persona is addressed. Upstream is referenced only by framework.skill: bmad-tea.

Nothing under frameworks/ is touched. Of ~40 fleet-wide murat hits, all but six are vendored and correctly keep the name. This PR changes only our layer — which is the point.

Addressing

address:
  handle: petry-projects/qa-lead
  aliases: []

The handle is an org team, never a user account. @qa-lead is a real, live GitHub account owned by someone unrelated to us — a workflow keying on the bare role would fire correctly, look healthy, and notify a stranger on every mention from a fleet of bots across public repos.

This PR declares the addressing contract but does not serve it: there is no router and no runtime yet, so nothing dispatches on the mention surface. qa-lead stays status: draft. The separation is deliberate.

Validator

Two new cross-invariants the JSON schema cannot express:

• address.handle's team slug == id — so the router resolves a mention by stripping the org prefix, and the role name is written exactly once.
• Handles and aliases are unique across ALL personas (case-insensitive). A collision is not cosmetic: the router resolves a handle to exactly one persona, so two claimants means a dropped mention or a double-fire. Aliases share the namespace with handles, so a rename's leftover alias cannot shadow a live role.

Live team properties (exists / privacy: closed / notifications_disabled) are deliberately not checked here — they need the network, and this validator is hermetic with a hermetic bats suite. Those belong to a CI step and the §7 DoD.

Verification

• bats tests/test_validate_personas.bats — 16/16 pass (10 pre-existing + 6 new: slug==id, handle collision, alias collision, case-insensitivity, distinct handles, no-address personas).
• validate-personas.py against the §4.1 schema: OK, invariants hold.
• python3 evals/validate-cases.py — OK, 5 skills / 46 cases, no cross-split id overlap. Eval ids renamed murat-* → qa-lead-*; only ids and descriptions changed — every input/expected is byte-identical, so the held-out set tests exactly what it did before.
• markdownlint-cli2 clean.
• holdout-guard passes: it fails only PRs authored by github-actions[bot] (the skill-proposer). This PR still needs CODEOWNER review over evals/ — that gate is intact.

Notes for reviewers

Merge order: .github#752 must merge first — validate-personas.py fetches the schema from .github@main, and address fails main's schema until then.

Deliberate follow-up: prompts/bmad/scrum-master.md still opens with "You are Bob" — the same §1.6 violation. That prompt drives the live initiative-planner, so changing its persona-adoption text is a behavior change that deserves its own PR and review, not a rider on a rename. The ("Murat") prose references in that file are fixed here, since they are pure prose with no runtime effect.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[coderabbitai]

Warning

Review limit reached

@don-petry, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 30 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 84d856e6-c730-4bae-a422-845111dab982

📥 Commits

Reviewing files that changed from the base of the PR and between 6effbe7 and 68c389d.

📒 Files selected for processing (12)

• docs/initiatives/idea-to-initiative-pipeline.md
• evals/murat/dev/cases.jsonl
• evals/murat/holdout/cases.jsonl
• evals/qa-lead/dev/cases.jsonl
• evals/qa-lead/holdout/cases.jsonl
• personas/murat/README.md
• personas/qa-lead/README.md
• personas/qa-lead/persona.yml
• personas/validate-personas.py
• prompts/bmad/README.md
• prompts/bmad/scrum-master.md
• tests/test_validate_personas.bats

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/qa-lead-role-rename

---

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.}

[gemini-code-assist]

Code Review

This pull request renames the "murat" persona to "qa-lead" across documentation, manifests, and evaluation files. It also introduces validation logic in validate-personas.py to ensure that persona addressing handles are unique and that their team slugs match the persona IDs. A review comment points out a potential IndexError crash in the validator if the handle is misconfigured without a / separator, suggesting a safer parsing approach.

[Copilot]

Pull request overview

This PR updates the worked-example persona from a person-name (murat) to a role-based, org-team addressable persona (qa-lead), and extends the persona validator + tests to enforce the new addressing invariants.

Changes:

• Add address.handle slug==id and global handle/alias uniqueness checks to personas/validate-personas.py, with new Bats coverage.
• Rename persona/eval content from murat-* to qa-lead-* (IDs, paths, docs), and introduce the qa-lead persona README/manifest content.
• Update BMAD orchestration and initiative pipeline docs to reference bmad-tea/qa-lead instead of the upstream person-name.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/test_validate_personas.bats	Adds hermetic tests for addressing invariants (slug==id, uniqueness, case-insensitive collisions).
personas/validate-personas.py	Implements new cross-invariants: handle slug==id and global handle/alias uniqueness.
personas/qa-lead/persona.yml	Renames the persona to qa-lead and adds the address contract block.
personas/qa-lead/README.md	New role-based persona documentation and addressing guidance.
personas/murat/README.md	Removes the old murat persona README (rename/move).
evals/qa-lead/holdout/cases.jsonl	New holdout eval cases with qa-lead-* IDs.
evals/qa-lead/dev/cases.jsonl	New dev eval cases with qa-lead-* IDs.
evals/murat/holdout/cases.jsonl	Removes old holdout eval cases under murat (rename/move).
evals/murat/dev/cases.jsonl	Removes old dev eval cases under murat (rename/move).
prompts/bmad/scrum-master.md	Updates prose references to the upstream agent skill id / qa-lead persona.
prompts/bmad/README.md	Updates vendored skill documentation to reflect bmad-tea behind qa-lead.
docs/initiatives/idea-to-initiative-pipeline.md	Updates initiative-planner documentation to reference bmad-tea / qa-lead.

[don-petry]

Dev-Lead — waiting on PR blockers (intent: review-changes)

PR: #1279
No changes were committed, but the PR still has blocking checks or reviews (failing or cancelled checks, or changes-requested reviews). The retry cron will re-attempt automatically. Next attempt after: 2026-07-17T02:43:21Z

[don-petry]

Note

@don-petry I reviewed this PR and no code changes were needed, but it still has blocking checks or reviews (failing or cancelled checks, or changes-requested reviews), so I cannot mark it done yet. I'll re-check automatically.
Next attempt after: 2026-07-17T02:43:21Z

[don-petry]

Addressed the review findings in 9d1411a.

Gemini (high) + Copilot — split("/", 1)[1] IndexError; .lower() on a non-string. Both are real, and both are unreachable in CI: schema validation runs before check_invariants, and the canonical schema's pattern — now ^petry-projects/[a-z0-9]+(-[a-z0-9]+)*$ after #752's own review tightened it — guarantees a well-formed string. Fixed anyway, because the validator accepts --schema: a test double or a stale local copy lets a malformed handle through, and this module's documented contract is "non-zero with a diagnostic", not a traceback.

• New handle_slug() validates org/team-slug and returns the slug, failing cleanly on a non-string or slash-less value.
• check_invariants now also validates address is a mapping and aliases is a list of well-formed handles. Alias shape is checked there, not in check_handles_unique — since that runs only after every manifest has cleared check_invariants, it can now treat handles and aliases as well-formed by construction. Noted in its docstring.

Copilot — capitalization. Correct; my rename sed introduced it. Only that one instance followed a period (the other three follow semicolons and were already right). Description text only — no input/expected touched, so the held-out set still tests exactly what it did before.

Verification

• bats tests/test_validate_personas.bats — 20/20 (16 + 4 new: bare handle, non-string handle, malformed alias, non-list aliases; each asserts a clean diagnostic and the absence of a traceback).
• validate-personas.py against the live merged schema on .github@main: OK, invariants hold.
• evals/validate-cases.py: OK, 5 skills / 46 cases.

[donpetry-bot]

Automated review — APPROVED ✓

Risk: MEDIUM
Reviewed commit: 3748c6b676bcc3a8576369055a6106debb911a77
Cascade: triage → deep (triage: haiku 4.5 → deep: opus 4.8 + duck: o4-mini → audit: fable 5)

Summary

Renames the murat worked-example persona to the role-based, @-addressable qa-lead and adds two hermetic cross-invariants (handle slug==id, handle/alias global uniqueness) to validate-personas.py with 6 new passing bats tests. The IndexError/AttributeError flagged by triage/gemini/copilot in the new validator code is an unreachable false positive: the .github@main schema pins handle to a string pattern requiring 'petry-projects/' and aliases to a string array, and fail() sys.exit(1)s on the first schema violation before check_invariants and check_handles_unique ever run, so every value reaching split('/')[1]/.lower() is a well-formed string. No secrets, no downstream consumers (none), and all CI gates (validate-personas, gitleaks, CodeQL, SonarCloud, agent-shield, holdout/test-deletion guards, 16/16 bats) are green.

Findings

• MINOR: check_invariants does address['handle'].split('/', 1)[1] with no guard. Flagged HIGH by gemini and copilot as an IndexError crash path, but it is unreachable: the persona schema pins handle to pattern ^petry-projects/[a-z0-9]+(-[a-z0-9]+)*$ (mandates a '/'), and main() runs jsonschema iter_errors -> fail() -> sys.exit(1) before check_invariants, so any handle reaching the split contains a '/'. Even if reachable, the crash is a non-zero exit of a hermetic CI lint that still rejects the bad manifest — no bypass. Optional defensive '/'-check would only improve the diagnostic message. (personas/validate-personas.py:125)
• MINOR: check_handles_unique calls handle.lower() and iterates address.get('aliases', []) without type-guarding (copilot: AttributeError on non-string handle / non-list aliases). Also unreachable: check_handles_unique runs only after the per-manifest loop, and fail()/sys.exit(1) fires on any schema violation first; the schema types handle as string and aliases as array-of-string, so loaded[] only ever holds schema-valid manifests. Same non-blocking diagnostic-quality nit. (personas/validate-personas.py:151)
• INFO: Copilot grammar nit: an eval case description starts mid-sentence with 'the QA Lead' after a period; should be 'The QA Lead'. Description text only — no effect on eval scoring (input/expected are byte-identical to the murat originals; holdout-guard and test-deletion-guard pass). (evals/qa-lead/holdout/cases.jsonl:3)
• INFO: Merge-order dependency noted by author: .github#752 (schema + standard) must merge before this PR, since validate-personas.py fetches the address schema from .github@main. The schema at main already carries the address block (verified during review), so the dependency appears satisfied. mergeStateStatus BLOCKED reflects REVIEW_REQUIRED only, not a failing check.

---

Reviewed by the PR-review cascade (triage: haiku 4.5 → deep: opus 4.8 + duck: o4-mini → audit: fable 5). Reply if you need a human review.

[don-petry]

Dev-Lead — review-changes (applied)

Changes committed and pushed.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[don-petry]

Dev-Lead — waiting on PR blockers (intent: review-changes)

PR: #1279
No changes were committed, but the PR still has blocking checks or reviews (failing or cancelled checks, or changes-requested reviews). The retry cron will re-attempt automatically. Next attempt after: 2026-07-17T03:11:00Z

[don-petry]

Note

@don-petry I reviewed this PR and no code changes were needed, but it still has blocking checks or reviews (failing or cancelled checks, or changes-requested reviews), so I cannot mark it done yet. I'll re-check automatically.
Next attempt after: 2026-07-17T03:11:00Z

[donpetry-bot]

Automated review — APPROVED ✓

Risk: MEDIUM
Reviewed commit: 68c389db9ffaffed9a33b77fdea93204d378bef3
Review mode: triage-approved (single reviewer)

Summary

Renames the worked-example persona murat → qa-lead (role, not person) per persona-standards §1.6, adds an org-team addressing contract (§4.1), and extends validate-personas.py with two cross-file invariants (handle slug == id; fleet-wide handle/alias uniqueness, case-insensitive) plus 11 new hermetic bats tests. Companion to petry-projects/.github#752. All prior review findings (Gemini high IndexError, Copilot type-safety and capitalization) are verifiably fixed at head.

Linked issue analysis

No linked issues (closingIssuesReferences is empty). The PR is a companion to petry-projects/.github#752 (the standard + schema), which is already merged — the validate-personas CI job passes against the live canonical schema.

Findings

Confirmation review of the triage assessment — the triage tier's low-risk clearance holds, with one nuance:

• Prior findings all resolved at head. Gemini's high-priority finding (split("/", 1)[1] IndexError on a slash-less handle) and Copilot's two findings (.lower() on non-string; capitalization in an eval description) were fixed in 9d1411a: the new handle_slug() helper validates isinstance(str) and "/" in value before splitting and emits a fail() diagnostic, check_invariants validates address mapping/alias shapes, and 4 new bats tests assert diagnostics without tracebacks. bats: green in CI.
• One review thread is technically unresolved (Gemini's IndexError thread) but it is substantively addressed: the fix is in the head commit with test coverage, and the author replied in-thread with specifics. Gemini threads are not auto-resolvable by the fix author; not treating this as blocking.
• Validator logic is sound. check_handles_unique runs only after every manifest clears check_invariants, aliases share the namespace with handles, and collisions are case-insensitive — matching how GitHub resolves team mentions. Ordering invariant is documented in the docstring.
• Rename hygiene is good. persona.yml moved via git mv (62% similarity, history preserved); eval case ids renamed with content otherwise unchanged (holdout-guard green); frameworks/ untouched as claimed — vendored upstream names correctly retained; no thin caller stubs modified.
• Addressing contract declared, not served — no router/runtime exists, persona stays status: draft, address.handle is an org team (never a user account, avoiding the real @qa-lead GitHub user). Live team properties are explicitly deferred to a networked CI job; validator stays hermetic. Sensible separation.
• Secret scan: the run_secret_scanning MCP tool is not available in this environment; the gitleaks CI check passed and the diff contains no credential-like content.

Risk is MEDIUM (non-trivial validator logic + cross-module rename), not HIGH: no auth/secrets/migrations, no workflow files touched, no standards violations.

CI status

All checks green for head 68c389d: shellcheck, ShellCheck, actionlint, bats, unit-tests, validate-personas, holdout-guard, gitleaks, CodeQL (actions + python), Agent Security Scan, agent-shield, SonarCloud quality gate, caller-stub-freeze, template-drift, gh-aw-compile, CodeRabbit. A few CANCELLED entries (review/review, dev-lead/dispatch, dev-lead/ci-relay) are concurrency-superseded duplicate runs — each has a later SUCCESS (ci-relay intentionally skipped via [skip ci-relay]). Dependency-audit jobs skipped (no matching ecosystems changed).

---

Reviewed automatically by the PR-review agent (single-reviewer mode: fable 5). Reply if you need a human review.