fix(query): normalize terminal workflow language

[mohanagy]

Summary

• normalize only complete, unquoted natural-language finished report and done report workflow terminals
• preserve protected-base behavior for identifiers, active operations, compound terminals, and locator precedence
• classify only the observed pipeline-orchestration recovery wording as a workflow
• record immutable beta.6 publication and manual-qualification failure history, plus the exact unpublished corrective package receipt

Core Reset contract

• Target: protected next only
• Parent: 66b795e76a76c6946b85e5eb878e3f576b4e3dbb
• Candidate head: 8c21c687adce70a483073c5406e0e5b1356cfaea
• Candidate tree: 5ffed3bc25cb537aafd3bb905ed9df08a7143d69
• Production delta: +68/-68/net 0, one existing source file
• Replacement: 1,424 source LOC / 60,995 emitted bytes
• Package: 102 files / 155,330 packed / 654,008 unpacked bytes
• No dependency, lockfile, package-version, graph/index/schema, CLI, or MCP change
• No npm publication, stable/latest change, GitHub Release, Registry metadata, tag, or main
• [P0][Quality]: Gate no-fallback retrieval against 0.32, beta.4, pinned Graphify, and control #631 remains open; this PR does not claim the final comparator or zero-fallback qualification

Verification

• Node 22 coverage: 83 files / 957 tests; all thresholds pass
• Node 20: 83 files / 957 tests, typecheck, build, build:eval, isolation pass
• Exact frozen GoValidate prompt, done, five controls, and recovery: ready 7/7 with identical 9-file / 12-excerpt / 20-hop evidence
• 516-question differential: zero unexpected planner differences
• Frozen issue [P0][Retrieval]: Plan obligations and return strict answer-ready workflow dossiers #630 benchmark: 14/14 ready; 100-sample warm p95 3.55 ms
• Packed CLI/application/MCP parity passes
• Registry validation, release hygiene, audit, eval regression, baseline contract, isolation, and diff hygiene pass

Relates to #631.

Summary by CodeRabbit

• Bug Fixes

  • Improved package receipt selection for terminal-language corrective releases.
  • Enhanced query planning for finished reports, compound terminals, workflow questions, and recovery scenarios.
  • Improved retrieval coverage for report-generation queries and broad prompts.

• Documentation

  • Updated beta.6 release records, roadmap, scorecard, and design documentation with publication details and qualification status.
  • Clarified that beta.6 failed manual qualification and that subsequent release promotion remains restricted.

• Tests

  • Expanded coverage for query planning, package parity checks, governance records, and retrieval obligations.

[coderabbitai]

📝 Walkthrough

Walkthrough

The change refines workflow query planning, records beta.6 publication and qualification outcomes, defines an unpublished terminal-language corrective candidate, and prioritizes that candidate in package measurement and replacement receipt resolution.

Changes

Query planning normalization

Layer / File(s)	Summary
Workflow and report query planning src/domain/query/plan.ts, tests/unit/query-plan.test.ts, tests/unit/retrieve-v2-contract-gaps.test.ts	Internal planning helpers are renamed. Orchestrated-flow intent and finished/done report terminal handling are added. Tests cover workflow obligations, compound terminals, recovery queries, and locator precedence.

Beta.6 corrective governance

Layer / File(s)	Summary
Publication records and corrective contract docs/core-reset/*, docs/designs/2026-07-19-core-reset.md, docs/roadmap.md, tests/unit/core-reset-governance.test.ts	Governance records beta.6 as published with failed installed-package qualification. The terminal-language corrective candidate includes measurements, parity, frozen-graph evidence, source scope, and publication restrictions.
Corrective receipt resolution .github/scripts/verify-packed-retrieval-parity.mjs, tools/eval/core-reset/verify-isolation.mjs, tests/unit/benchmark-suite-isolation-docs.test.ts	Receipt lookups prefer terminal_language_corrective measurements before existing corrective fallbacks. The parity test checks the new candidate path.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

• mohanagy/madar#579: Introduced the Core Reset governance documents and tests extended here.
• mohanagy/madar#581: Introduced the isolation receipt fallback updated here.
• mohanagy/madar#638: Updated the packed parity receipt lookup extended here.

Suggested labels: core-reset

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the primary change: normalizing terminal workflow language in query planning.
Description check	✅ Passed	The description clearly covers behavior, scope, Core Reset history, issue status, and verification results, despite missing some template headings and checkboxes.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch agent/issue-631-terminal-language

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/domain/query/plan.ts`:
- Line 196: Restrict the recovery predicate in the query classification logic to
match only the canonical observed pipeline-orchestration wording, rather than
any “where is/are/was/were … flow|workflow|pipeline orchestrated” variant.
Preserve the existing locator classification precedence for all other query
forms and leave downstream retrieval behavior unchanged.

In `@tests/unit/core-reset-governance.test.ts`:
- Around line 2626-2631: Strengthen the terminal corrective measurement
assertions in the test around TERMINAL_LANGUAGE_BASE: recompute both production
and test diffs, verify the allowed paths and --numstat values match the
manifest’s 68/68/0 measurement and test_sources allowlist, and compute the
binary diff SHA-256 to compare with diff_sha256. Keep the existing
src/domain/query/plan.ts path assertion, but ensure stale or unrecorded changes
cannot satisfy terminal_language_corrective.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: edb8829a-1f2d-40be-a01a-bcc1f803ee4a

📥 Commits

Reviewing files that changed from the base of the PR and between 66b795e and 8c21c68.

📒 Files selected for processing (11)

• .github/scripts/verify-packed-retrieval-parity.mjs
• docs/core-reset/removal-manifest.yml
• docs/core-reset/scorecard.md
• docs/designs/2026-07-19-core-reset.md
• docs/roadmap.md
• src/domain/query/plan.ts
• tests/unit/benchmark-suite-isolation-docs.test.ts
• tests/unit/core-reset-governance.test.ts
• tests/unit/query-plan.test.ts
• tests/unit/retrieve-v2-contract-gaps.test.ts
• tools/eval/core-reset/verify-isolation.mjs

[mohanagy]

Exact-head independent-review and merge-gate receipt

• Target: protected next; current protected base 66b795e76a76c6946b85e5eb878e3f576b4e3dbb, tree ff57246a55e51459836946d6b5b9853d0cdcc372
• Reviewed head: 8c21c687adce70a483073c5406e0e5b1356cfaea
• Reviewed tree: 5ffed3bc25cb537aafd3bb905ed9df08a7143d69
• Sole parent: 66b795e76a76c6946b85e5eb878e3f576b4e3dbb
• Exact-head CI: all six required jobs passed in https://github.com/mohanagy/madar/actions/runs/30753189853
• CodeRabbit: status passed. Its two actionable threads were verified against the contract, answered with evidence, and resolved without changing the reviewed head; zero unresolved review threads remain. The walkthrough also reports a non-blocking docstring-coverage warning, not a required repository gate.
• Independent semantic review: no blocker; 39 adversarial terminal and intent cases had zero unexpected changes, with explicit identifier, active operation, compound terminal, final/complete, and locator controls preserved.
• Independent test/spec review: no blocker; 145 focused tests passed, the 516-question differential produced only the 96 authorized finished/done normalizations and 9 authorized recovery promotions, and the isolated packed real-GoValidate matrix was ready 7/7 with identical 9-file / 12-excerpt / 20-hop evidence.
• Independent governance review: no blocker; exact ancestry, source delta/hash, package receipt, protected-target boundary, isolation, parity, and non-publication constraints were verified.

The exact head is clean and mergeable. This receipt claims no npm publication, GitHub Release, Registry metadata, tag, stable/latest change, final #631 comparison, or main change. Issue #631 remains open.