docs(setup): add PR-management welcome guide

[leninworld]

This PR addresses part of issue #1655.

Summary

• Add the first-run welcome guide for the PR-management family: configuration, the queue workflow, example prompts and overrides. Every write it describes stays confirmed, and quick-merge leaves the merge command to the user.
• Link the guide from the family documentation, and cover its first-run sequence and shared footer in test_welcome.py.
• Add the rendered preflight-floor fixture case-11-pr-management-first-run and update the generated case-9 report and case counts. The guide uses the shared welcome section, like the utilities guide (PR #1669); no core logic changes.

Type of change

• Skill change (.claude/skills/<name>/) — eval fixtures updated below
• Tool / bridge contract (tools/<system>/*.md)
• Python package (tools/*/ with pyproject.toml)
• Groovy reference impl
• Cross-cutting (RFC, AGENTS.md, sandbox, privacy-LLM)
• Documentation (docs/, README.md, CONTRIBUTING.md)
• Project template (projects/_template/)
• CI / dev loop (prek, workflows, validators)
• Other: setup-preflight welcome guide, regression test and preflight-floor fixture

Test plan

Linux ARM64 Docker, Python 3.12, uv 0.11.8, Git 2.47.3.

• prek run --all-files passes (every hook, including workspace pytest and lychee)
• For Python packages touched: uv run pytest / ruff check / mypy passes (setup-preflight: 257 passed; mypy: no issues in 26 source files)
• For Groovy bridges touched: command-line invocation tested end-to-end
• For skill changes: eval suite passes for the affected skill
  (PYTHONPATH=tools/skill-evals/src python3 -m skill_evals.runner tools/skill-evals/evals/<skill>/)
• For skill behaviour changes: a new or updated eval fixture is included in this PR
  (a regression test for the bug fixed / the behaviour added — see CONTRIBUTING.md)
• Other: render-preflight-fixtures.py --check passes; the new sequence test fails without the guide and passes with it; python -m setup_preflight.welcome show --family pr-management renders the guide and shared footer (69 lines).

Not run: provider-driven skill evals (they need model calls) and hosted CI (not yet run). The container used a child-subreaper wrapper for normal orphan reaping; project code and test assertions are unchanged. A wheel-build probe fails identically on unchanged main and on this branch, because the package config adds setup_preflight/guides/_footer.md twice. No packaging change or successful wheel build is claimed.

RFC-AI-0004 compliance

• HITL — any new mutation is gated on explicit user confirmation
• Sandbox — no new unrestricted host access; network reach declared in the adapter
• Vendor neutrality — placeholders (<PROJECT>, <tracker>, <upstream>, <security-list>) used in all skill / tool prose (the check-placeholders prek hook is the mechanical gate)
• Conversational + correctable — agentic-override path documented if behaviour is adopter-tunable
• Write-access discipline — no autonomous outbound messages; drafts only, sent on confirmation
• Privacy LLM — private content does not reach a non-approved LLM; redactor invoked where needed

No new mutation, host access, outbound write or LLM hop; the guide describes the family's existing confirmation gates.

Linked issues

Refs #1655

Notes for reviewers (optional)

• Case 11 follows the requested contiguous numbering on current main. PR #1673 independently adds another case 11; whichever lands second will need renumbering, fixture regeneration and count updates.

• If you would prefer a per-family sections/welcome-pr-management.md (asked on the issue), I can add it.

• Review corrections: explicitly use pr-triage's surface hash in the fixture, provide nine unique skill prompts, match the other guides' heading accents, and rename the shipping test to test_the_rules_section_and_all_guides_ship. The follow-up also repairs invalid JSON in the applied state.json suggestion. A narrow spell-check exclusion covers SHA-256 fixture hash values.

Validation evidence

---

Was generative AI tooling used to co-author this PR?

• Yes — Codex (GPT-6), with independent review by Claude (Opus 5.5)

Generated-by: Codex (GPT-6)

The contributor has read every line of the diff and can answer review questions about it.

[Kaap10]

Thanks @leninworld for putting this together! The structure and test coverage follow the pattern established in #1653 and #1669 well.

Before merge, there are a few fixture state corrections and consistency adjustments needed:

1. surface_hash in eval fixture: report.md quotes sha256:2edc90a1b7f3440d (the hash for list-skills), whereas pr-triage is sha256:54dc4ba1be7b36d4.
2. Fixture sequence numbering: There is no case-11 on main, so naming this case-12 creates an unexpected gap.
3. Full skill prompt coverage: Include prompt examples for all 9 skills under ## Just ask (quick-merge, reviewer-routing, stats).
4. Visual parity: Add standard emoji accents (🐦, 🔄, ⚙️, 💬, 🎨) to match the other family guides.

[leninworld]

Thanks @leninworld for putting this together! The structure and test coverage follow the pattern established in #1653 and #1669 well.

Before merge, there are a few fixture state corrections and consistency adjustments needed:

1. **`surface_hash` in eval fixture**: `report.md` quotes `sha256:2edc90a1b7f3440d` (the hash for `list-skills`), whereas `pr-triage` is `sha256:54dc4ba1be7b36d4`.

2. **Fixture sequence numbering**: There is no `case-11` on `main`, so naming this `case-12` creates an unexpected gap.

3. **Full skill prompt coverage**: Include prompt examples for all 9 skills under `## Just ask` (`quick-merge`, `reviewer-routing`, `stats`).

4. **Visual parity**: Add standard emoji accents (`🐦`, `🔄`, `⚙️`, `💬`, `🎨`) to match the other family guides.

Done. Thank you!

[Kaap10]

LGTM! Thanks @leninworld for addressing all review points so quickly.

Next maintainer will look into it and take it forward!