fix: reconcile role handoff and validation completion

[jamesreimer]

Role boundaries were ambiguous after generic continuation instructions, and an implementation could be returned for review before required validation finished. This change makes role continuity, explicit transfer, reviewer non-mutation, source verification before mutation, and final-success validation explicit in the canonical workflows. It also pins Publication and Release Integrity (PRI) and connects publication evidence to its applicable requirements.

Authority: Agent Workflows #11, settled Planning decision, and the bounded Implementation request. This PR does not close #11. Completion is return to Planning / Primary Author for exact-candidate review and independent-review disposition.

Implementation → review handoff

Candidate: 5bd3e9fcf639f8d212d9b476226e5b0118777d3a
Comparison base for the bounded correction: 2b6f1fa73afeb458bcb19650c546de72b13238fc
Original PR base: 2d571bd7073752e2681b1d25201f42c714ffbd9b.

The current candidate is one correction commit beyond the independently reviewed candidate: fix: preserve PRI guidance status and exposure triggers. That correction changes exactly roles/publication.md, handoffs/review-to-publication.md, and workflows/reviewed-change.md.

• F-1: source-derived PRI requirements are no longer restated as workflow-local guidance in workflows/reviewed-change.md. The local section references the applicable source-derived guidance and retains no redesign, no silent substitution, and correction routing through the authorized role handoff.
• F-2: observable publication state is accounted for before replacing, retrying, withdrawing, correcting, or claiming completion in the source-derived passages of roles/publication.md and handoffs/review-to-publication.md (PRI §§4, 13).

Planning accepted F-1 and F-2 in Issue #11 and has accepted the bounded source corrections. N-1, N-2, and N-3 remain accepted non-blocking notes with no required change. No other semantic changes were introduced by the correction.

Planning / Primary Author: arrange the required bounded independent re-review under the settled contract. The independent reviewer should inspect the immutable candidate against the bounded correction comparison base, verify F-1/F-2 and the exact-candidate validation evidence below, and return a supported conclusion and all material finding dispositions to Planning in this PR. Review must not mutate the candidate. Route any corrections through authorized Implementation; no merge, publication, tag, release, consumer propagation, Standards Templates modification, or issue closure is authorized by this handoff.

Requirement mapping and changed files

Requirement	Implemented artifacts
Role continuity; explicit reassignment or instantiated transfer; generic proceed/continue	roles/planning.md, roles/implementation.md, roles/independent-review.md, roles/publication.md, workflows/reviewed-change.md
Ready-to-use next-role handoff; Planning does not implement	roles/planning.md, all role continuity paragraphs, handoffs/return-of-control.md, workflows/reviewed-change.md
Reviewer does not mutate; correction routes to authorized implementor	roles/independent-review.md, handoffs/implementation-to-review.md, handoffs/review-to-correction.md, workflows/reviewed-change.md
Required validation final success; pending/queued/running/cancelled/skipped/unresolved are not success; open PR is insufficient; final non-success stays incomplete	roles/implementation.md, handoffs/implementation-to-review.md, handoffs/review-to-correction.md, workflows/reviewed-change.md
Exact approved-source correspondence before mutation; labels insufficient; SAP semantics	roles/implementation.md, handoffs/planning-to-implementation.md, handoffs/review-to-correction.md, workflows/reviewed-change.md
PRI pin, publication authority/correspondence/identities/resulting-state verification/partial exposure; no redesign or substitution	GOVERNING-SOURCES.md, roles/publication.md, handoffs/review-to-publication.md, workflows/reviewed-change.md

These are all 11 changed files relative to the original PR base; the bounded correction changes only the three files identified above. Roles repeat the continuity rule so each selected role carries its boundary when used independently. No validator, runtime, schema, repository-local consumer command, or organization-neutral standard was added.

Governing sources and consumer impact

Added PRI edition 1.0, pinned to Standards Templates debd50926ea45c371a7b10237ef9aba1b3b04304. Read its source standard and README directly at that revision; the README confirms edition 1.0. Existing five pins remain at df73ad31cdbb15210ad9670d669d3c567a79a1d3. Rechecked SAP §§5–8 against its pinned text for identity and working-state correspondence.

PRI guidance references its sections rather than copying normative text. New role/validation/transfer mechanics remain under workflow-local headings; PRI-derived guidance remains source-derived and conditional on actual project applicability. SAP still owns provenance, OEC execution authority, and downstream adoption remains separate. No universal platform, four-actor mandate, mandatory retention scheme, or adoption is introduced.

Consumers deliberately select the new revision and map actual governing authority; there is no propagation. The publication handoff now asks for authorized source state, identity meanings and surfaces, correspondence, completion verification, and applicable prior-exposure assessment. Return-of-control adds a next-role instruction field.

Validation and evidence

Current candidate 5bd3e9fcf639f8d212d9b476226e5b0118777d3a: the complete local pre-commit qualification was rerun and passed, including the regression suites and artifact checks; only the expected no-files symlink hook skipped. git diff --check passed, npm audit reported zero vulnerabilities, and the working tree was clean. The following original implementation details are retained as historical evidence; the hosted validation entry identifies the current candidate.

• .venv/bin/pre-commit run --all-files --show-diff-on-failure: final PASS across all applicable hooks, including Markdown/link regression suites and artifact tests. Broken-symlink hook reports the expected no-files skip. Initial run was superseded by a formatting cleanup; the complete staged final tree passed on rerun.
• git diff --check and git diff --cached --check: PASS.
• python3 scripts/validate_local.py: PASS.
• python3 -m unittest discover -s tests: PASS, 5 tests.
• npm audit: PASS, 0 vulnerabilities.
• Local Node: v24.18.1. The validated staged tree was committed without further content changes; worktree is clean.
• Hosted required Repository validation: final SUCCESS on exact candidate 5bd3e9fcf639f8d212d9b476226e5b0118777d3a; run 35961031094, job and logs. The run is completed with conclusion success and its head SHA matches the candidate. The workflow checks out the exact PR head. Live main rules require this check; gh pr checks 12 --required reports pass.
• PR head read back matches the current candidate above. The bounded correction comparison base and original PR base are recorded separately. Candidate is fully validated and ready for bounded independent re-review; control returns to Planning / Primary Author.

Manual implementor inspection: checked the requirement mapping above against the diff; checked role/handoff/composition consistency; verified new source citations against PRI and SAP; checked conditional applicability and source/workflow-local separation. This is not independent review.

Validation limitation: CONTRIBUTING explicitly limits the local validator to required artifacts and obvious live identifiers and excludes workflow-semantic parsing. Existing tests cover that guard, Markdown authoring, and local links. No prose-matching semantic tests or new parser were added. Semantic correctness and source-force interpretation require the requested independent review.

Deviations, findings, and unresolved conditions

No Planning deviations or architectural conflicts found. No unresolved implementation conditions. The validator's intentional semantic-coverage limit is recorded above. The prior independent review returned ACCEPT WITH MINOR CORRECTIONS; Planning accepted F-1/F-2 and subsequently accepted their bounded source corrections. Bounded independent re-review and its disposition remain outstanding; publication requires separate authorization. The working tree was clean at the original PR base; the full PR changes the 11 mapped files, and the correction changes exactly the three files identified above.