Simplify wo-pr and seda-pr into outcome-led PR workflows

[mosobande]

Stack: ori → #25 → #98 → #99 → #100. This layer targets #98 (feat/html-artifact-base-template). Review its diff against that parent; merge bottom-up. #95 is excluded. Existing commits are preserved with a merge commit; no force-push or PR merge performed.

Outcome

Make PR babysitting an inspect → correct → publish → wait loop, with stack guidance loaded only when dependencies matter. Publish one accurate PR without routine provider/Git instruction duplication.

Changes

• Explicit babysit/get-ready requests authorize scoped corrections; watch/status/read-only stays read-only.
• Source fixes still use alaga for implementation, proof, and review; wo-pr retains the loop and granted scope.
• Preserve exact head/base evidence, provider trust/readback, retry limits, retarget/rewrite boundaries, and ordered stack reconciliation.
• Resolve the contradictory blanket triage/current-head feedback guidance; revalidate older still-applicable findings.
• Preserve provider-readiness identifiers for callers; return concise result-sized reports.

Proof

Package and plugin validators pass; 21 repository contract tests pass. Eight same-scenario fresh-context baseline/candidate simulations and independent static review preserve the relevant authority, stack, stale-base, and ambiguous-write decisions. Source-correction routing was clarified and rechecked afterward. These are simulated decisions, not live GitLab/Claude/provider mutation proof.

Ordinary loading-path words: wo-pr 1,723 → 538; seda-pr 1,392 → 472. The duplicate repository evaluation report has been removed; this PR discussion is the evidence record. Cleanup changes documentation only; workflow behavior is unchanged from the reviewed candidate. No runtime scripts or standing prompt assertion harness added.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}