[taras]

Why

Retire the automated PR reviewer at the maintainer's request. It repeatedly proposed deleting upstream patch context, lint directives and required provenance comments, creating review work without demonstrated correctness findings in the recent sample. Fixes #1226.

What changes

• Remove PR Review and its companion manual Repo Analysis workflow, the code-review-agent workspace, .reviews entrypoints and policies, and their dedicated specifications/tests.
• Remove review/analyze task aliases, package dependencies and lock entries, publication jobs, documentation/build registration, configuration exemptions and current feature references.
• Preserve general plugin loading, reserved-component precedence, origin/digest, compiled-asset and testing-host coverage using the existing small declaring.mjs fixture. Reviewer-only tests are deleted.
• Keep the existing blocking lint policy and tests, documented in specs/oxlint-policy-spec.md. Normal CI and the implementation workflow's independent Fable/Astra reviews remain.
• Remove the obsolete DeepInfraProvider example from the Sample diagnostic.

Both hosted workflows were disabled in GitHub before publication. There are no running or queued PR Review jobs. Historical reviews, credentials and recorded test timings are retained; deleted-file timing entries are ignored by the existing measurement contract.

Review guide

Start with the workflow/package removals and workspace locks. Then inspect scripts/validate-documentation.ts, publication generation, declaring.mjs, and the plugin/compiled-plan test migrations. The source deletion accounts for most of the diff: 115 files, +309 / -11,410. No replacement reviewer or public API is added.

How to verify it

Independent Fable review passed on exact final head 6e2dd3df3385e29bac309c6094c08889def9a5ec, after reviewing the retirement on parent 99b22b60 and then the two-line Sample diagnostic delta in the same reviewer session.

Focused evidence:

• Core schemas/source transforms, package inventory, lint-policy and manifest/lockstep checks pass.
• Plugin host/CLI, documentation, publication/CI workflows, testing-host and six Markdown-tier suites pass. Making the fixture's Stamp registration non-reserved causes the expected precedence test failure; the fixture was restored.
• Fresh deno task setup, deno task build, the actual plugin/compiled-plan suites and npm CLI package suite pass on 99b22b60.
• Repository lint, formatting, dead-code and affected TypeScript checks pass. The final diagnostic-only successor additionally passes Sample tests (5 groups / 25 steps), file typecheck/lint/format and diff checks. Dependency and plugin-boundary bytes are unchanged by that successor.
• deno task verify:clean: PASS on exact final commit 6e2dd3df, including clean setup, offline builds/release compilation, resolution and Deno/Node/Bun interference proof with unchanged tracked files, node_modules and lockfile.

The first dependency/network, npm-cache permission and sandboxed process-inspection failures are preserved. Setup recovered with an owned cache and a fresh prepared layout. The final verification runs outside the sandbox so its existing process-join step can inspect its own children. No preflight, assertion, deadline or signing rule was bypassed. Required CI and current main health remain delivery gates.

Scope

The existing lint engine and its resolved optional peer remain pinned; this retirement performs no unrelated dependency upgrades. Shared runtime APIs keep their existing consumers and contracts. Original user changes, #1203 and the two planned pilots are untouched.

Scope confirmation

• All changes support retirement or preserve its shared consumers.
• No new abstraction, review framework or compatibility facade.
• Generated publication workflow and all workspace locks reflect the removal.
• Review and verification evidence identify the source revision they actually tested.