the-intent-audit-says-audit-abcd-intent-review-becomes-abcd

Summary

Delivers adr-40's first named rename: abcd intent review → abcd intent audit (and review ingest → audit ingest), the intent-fidelity-reviewer agent → intent-auditor, the intent_review task-class token → intent_audit. Clean break, no aliases. Ships only after spc-27's extended surface_coverage is armed; the table row flips in the same change, so the gate proves the migration complete.

Scope — the live sweep (verified inventory, 2026-08-16)

Code and surface:

• internal/core/intent/review.go (+ review_test.go) → audit.go / audit_test.go; exported Review* identifiers → Audit*; the review_emit_error JSON key (ReviewEmitError, intent.go) → audit_emit_error; create.go's references follow.
• internal/surface/cli/cli.go — the review sub-command under intent → audit; help text; JSON rendering.
• agents/intent-fidelity-reviewer.md → agents/intent-auditor.md (name, frontmatter, prompt body, capability_scope.task_classes); agents/ README.md roster row.
• commands/intent.md — the Review/ingest section, argument-hint.
• docs/reference/cli/commands.md — regenerate.

Record (current-state surfaces only):

• Brief: 01-product/01-press-release.md, 01-product/03-mental-model.md, 02-constraints/04-naming.md (the task-class enum row: intent_review → intent_audit), 04-surfaces/README.md, 04-surfaces/05-intent.md, 04-surfaces/06-capture.md, 04-surfaces/09-reflect.md, 05-internals/README.md, 01-agents.md, 02-adapters.md, 03-configuration.md, 04-universal-patterns.md, 05-prompt-quality.md, 06-lint.md, 06-delivery/02-verification-matrix.md.
• The intent sub-verb table row (spc-27's format): review → audit, same change.
• CHANGELOG.md — breaking entry.

Not swept (historical, append-only, or dated — AC 3): ADRs, dated research notes, dated plans, resolved/wontfix issues, shipped/superseded intents, DECISIONS.md, review-store artefacts, git history. A test or grep in the PR description demonstrates remaining occurrences are all in these classes. Amendment (2026-08-26): intents/disciplines/ appeared in neither the swept inventory nor this list and was swept late (iss-2608261437043634) — it is live state with no reachable reconciliation event, since intent audit refuses a non-shipped intent; drafts/ and planned/ stay deliberately unswept under the iss-94 convention (each reconciles when next planned). Stored artefact formats (abcd-review: markers, review-<ts> receipt directories) keep their frozen spellings.

Approach

1. Precondition check first: assert spc-27's sub-verb check is armed (the extended rule enabled and the intent table present). If not armed, STOP — the ordering is the plan's non-negotiable constraint.
2. Rename code path (Audit, AuditIngest, audit_emit_error), watching the renamed tests fail under the old spelling first, then pass.
3. Rename the agent file and roster; the task-class token in the agent frontmatter and the naming registry move together (the schema cross-check test in the binary must be updated in the same commit — the enum is PR-to-extend and no shipped check reads stored artefacts, so previously ingested verdicts stay valid; assert by re-running the ingest parser over an existing stored verdict fixture).
4. Sweep the brief and command surfaces; flip the sub-verb table row.
5. Regenerate surface.json (go generate ./internal/surface/cli) and the CLI reference; record-lint, docs-lint, and the surface drift test exit 0 — that is the proof of completeness.
6. Old spelling behaviour: abcd intent review exits as an unknown sub-command (cobra default); a surface test pins this.

Acceptance-criteria mapping

• New spellings, old gone — steps 2, 6; surface tests for audit, audit ingest, and the unknown-command refusal of review.
• Live sweep total — the Scope inventory; step 5's gates prove it.
• Historical records untouched — the Not-swept list; PR shows the residual grep classified.
• Armed-gate ordering — step 1; the run stops if spc-27 is not armed.
• Stored artefacts still parse — step 3's fixture assertion.
• Breaking CHANGELOG — Scope; version derivation reads it.

Every behaviour change lands with a test watched fail before the change.