@holdyourvoice/hyv 3.5.1 → 3.6.0

package/Readme.md

@@ -131,7 +131,7 @@ producer | hyv final-check -
 | `hyv score <draft> <profile-v3.json> <heldout...>` | Measure the draft against a separate held-out local writing range; may abstain. |
 | `hyv ingest <source> <export> --owner=owner --output=/absolute/dir` | Create redacted, owner-authorized Gmail or Telegram samples and a text-free receipt. |
 | `hyv analyze <draft> <profile.json>` | Run VoiceDNA, AI Editor, and hygiene checks. |
-| `hyv strict-check <draft> <profile-v3.json> <sample...>` | Run the opt-in strict local quality gate. It requires a calibrated V3 profile built from five or more samples, five non-duplicate samples with a consistent visible format and 1,500 words total, and returns `strict-ready`, `needs-human-review`, or `blocked`. |
+| `hyv strict-check <draft> <profile-v3.json> <sample...>` | Run the calibrated local strict-quality gate. It requires a V3 profile built from five or more samples, five non-duplicate validation samples with a consistent visible format and 1,500 words total, and returns `strict-ready`, `needs-human-review`, or `blocked`. |
 | `hyv hygiene <draft> [--fix]` | Inspect hidden Unicode or write a conservative cleaned copy. |
 | `hyv inspect-hidden-text <draft> [policy.json]` | Inspect hidden text with an optional policy. |
 | `hyv apply-hidden-text-policy <draft> <policy.json> <output>` | Apply approved hidden-text removals. |
@@ -143,7 +143,7 @@ producer | hyv final-check -
 | `hyv fact-lint <draft\|-> --source=id:path` | Check claims against local source files. |
 | `hyv logic-lint <draft\|-> [brief.json]` | Check deterministic document logic. |
 | `hyv batch-analyze <draft...>` | Find repeated openings and endings across drafts. |
-| `hyv rewrite-prompt <draft> <profile.json>` | Create a constrained editing brief. |
+| `hyv rewrite-prompt <draft> <profile.json>` | Create a strict constrained editing brief. |
 | `hyv prepare-rewrite ...` | Create a fingerprint-bound edit task. |
 | `hyv apply-rewrite ...` | Apply and verify a response to an edit task. |
 | `hyv prepare-judgment ...` | Create a pre-edit or post-candidate judgment task. |
@@ -159,6 +159,8 @@ producer | hyv final-check -
 | `hyv agent list\|validate\|describe\|emit` | Inspect or emit portable agent contracts. |
 | `hyv mcp` | Start the local MCP server on standard input/output. |
 
+`rewrite-prompt`, `prepare-rewrite`, `apply-rewrite`, and `verify` are strict by default: every active AI Editor finding must be repaired, and verification rejects a candidate that leaves one unresolved. A Profile v3 `disabled` policy is the only deliberate exception. `strict-check` adds the separate calibrated voice-evidence decision.
+
 Most commands return JSON. Exit code `0` means the command completed, `2` means a content or policy gate failed, and `1` means the command or input was invalid.
 
 Run `hyv <command>` without enough arguments to see its exact usage. Read the [CLI reference](docs/wiki/CLI-Reference.md) for every option.

package/dist/cli.test.js

@@ -272,7 +272,7 @@ test('uses exit code 2 for a failed candidate gate and 1 for misuse', () => {
         assert.equal(run(['profile', profile, first, second, '--avoid=unlock']).status, 0);
         const verification = run(['verify', original, candidate, profile]);
         assert.equal(verification.status, 2);
-        assert.deepEqual(Object.keys(JSON.parse(verification.stdout)).sort(), ['candidate', 'finalOutput', 'logicLint', 'original', 'passed', 'preservationScore', 'regressions', 'version']);
+        assert.deepEqual(Object.keys(JSON.parse(verification.stdout)).sort(), ['candidate', 'finalOutput', 'logicLint', 'original', 'passed', 'preservationScore', 'regressions', 'strictFindings', 'version']);
         assert.equal(run(['unknown-command']).status, 1);
         assert.equal(run(['mcp', 'unexpected']).status, 1);
     }

package/dist/mcp.js

@@ -74,7 +74,7 @@ server.registerTool('hyv_analyze', {
     annotations: { readOnlyHint: true },
 }, async ({ draft, profile_json, writing_brief_json }) => guardedJson(() => analyzeForMcp(draft, profile_json, writing_brief_json)));
 server.registerTool('hyv_strict_check', {
-    description: 'Run the opt-in local strict quality gate. It requires a calibrated Profile v3 and local writing samples; it returns strict-ready, needs-human-review, or blocked without changing text or learning state.',
+    description: 'Run the calibrated local strict-quality gate. It requires a Profile v3 and local writing samples; it returns strict-ready, needs-human-review, or blocked without changing text or learning state.',
     inputSchema: { draft: writing, profile_json: profileJson, samples: strictSamples, writing_brief_json: writingBriefJson.optional() },
     annotations: { readOnlyHint: true },
 }, async ({ draft, profile_json, samples: localSamples, writing_brief_json }) => guardedJson(() => strictCheckForMcp(draft, profile_json, localSamples, writing_brief_json)));
@@ -128,7 +128,7 @@ server.registerTool('hyv_logic_lint', {
     annotations: { readOnlyHint: true },
 }, async ({ draft, writing_brief_json }) => guardedJson(() => logicLintForMcp(draft, writing_brief_json)));
 server.registerTool('hyv_rewrite_prompt', {
-    description: 'Create a constrained editing brief. Explicit local examples are redacted in memory and injected only as advisory cadence evidence. It does not rewrite the draft or call a model.',
+    description: 'Create a strict constrained editing brief. Every active AI Editor finding is a required repair; explicit local examples remain advisory cadence evidence. It does not rewrite the draft or call a model.',
     inputSchema: { draft: writing, profile_json: profileJson, writing_brief_json: writingBriefJson.optional(), examples: writingExamples.optional() },
     annotations: { readOnlyHint: true },
 }, async ({ draft, profile_json, writing_brief_json, examples }) => guardedJson(() => rewritePromptForMcp(draft, profile_json, {}, writing_brief_json, examples)));
@@ -138,12 +138,12 @@ server.registerTool('hyv_find_writing_examples', {
     annotations: { readOnlyHint: true },
 }, async ({ query, examples }) => guardedJson(() => findWritingExamplesForMcp(query, examples)));
 server.registerTool('hyv_prepare_rewrite', {
-    description: 'Prepare a local, versioned rewrite task. The caller may forward it to a provider; doing so shares the draft and must be an explicit choice.',
+    description: 'Prepare a strict local, versioned rewrite task. Every active AI Editor finding is eligible for source-faithful repair; the caller may forward the task to a provider only by explicit choice.',
     inputSchema: { draft: writing, profile_json: profileJson, copy_spec_json: copySpecJson.optional(), writing_brief_json: writingBriefJson.optional() },
     annotations: { readOnlyHint: true },
 }, async ({ draft, profile_json, copy_spec_json, writing_brief_json }) => guardedJson(() => prepareRewriteForMcp(draft, profile_json, copy_spec_json, writing_brief_json)));
 server.registerTool('hyv_apply_rewrite', {
-    description: 'Validate and apply a model response to a prepared task, then run the local gates. It never calls a provider or stores source or candidate text.',
+    description: 'Validate a model response to a prepared task, then reject any candidate with an unresolved active AI Editor finding. It never calls a provider or stores source or candidate text.',
     inputSchema: { task_json: z.string().min(1).max(250_000), response_json: z.string().min(1).max(100_000), profile_json: profileJson },
     annotations: { readOnlyHint: true },
 }, async ({ task_json, response_json, profile_json }) => guardedJson(() => applyRewriteForMcp(task_json, response_json, profile_json)));
@@ -164,12 +164,12 @@ server.registerTool('hyv_reduce_judgment', {
     annotations: { readOnlyHint: true },
 }, async ({ envelopes_json }) => guardedJson(() => reduceJudgmentForMcp(envelopes_json)));
 server.registerTool('hyv_verify', {
-    description: 'Verify a revised candidate against an original draft and portable profile without changing learning state.',
+    description: 'Verify a revised candidate against an original draft and portable profile. Every active AI Editor finding fails default verification; learning state remains unchanged.',
     inputSchema: { original: writing, candidate: writing, profile_json: profileJson, writing_brief_json: writingBriefJson.optional() },
     annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
 }, async ({ original, candidate, profile_json, writing_brief_json }) => guardedJson(() => verifyForMcp(original, candidate, profile_json, writing_brief_json)));
 server.registerTool('hyv_verify_copy_spec', {
-    description: 'Verify a candidate against the existing voice gates and a local CopySpec. Immutable claims remain verbatim unless atoms are supplied; then each declared atom must remain. Prohibited claims fail closed.',
+    description: 'Verify a candidate against strict default voice gates and a local CopySpec. Every active AI Editor finding fails verification; immutable claims remain verbatim unless atoms are supplied, and prohibited claims fail closed.',
     inputSchema: { original: writing, candidate: writing, profile_json: profileJson, copy_spec_json: copySpecJson, writing_brief_json: writingBriefJson.optional() },
     annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
 }, async ({ original, candidate, profile_json, copy_spec_json, writing_brief_json }) => guardedJson(() => verifyCopySpecForMcp(original, candidate, profile_json, copy_spec_json, writing_brief_json)));
@@ -250,7 +250,7 @@ if (redactsSensitiveInputs) {
         annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
     }, async (args) => guardedJson(() => prepareRebuildForMcp(args.draft, args.profile_json, args.reduction_json, args.copy_spec_json, args.capability_json, loadApprovalContext(), args.writing_brief_json, args.recomposition_policy_json), 'Rebuild preparation failed.'));
     server.registerTool('hyv_apply_rebuild', {
-        description: 'Validate and evaluate a whole-document rebuild response against a prepared authorized rebuild task. Capability input requires host-guaranteed sensitive-input redaction. It never calls a provider.',
+        description: 'Validate a whole-document rebuild response against a prepared authorized rebuild task and reject unresolved active AI Editor findings. Capability input requires host-guaranteed sensitive-input redaction. It never calls a provider.',
         inputSchema: { task_json: lifecycleJson, response_json: z.string().min(1).max(100_000), profile_json: profileJson, capability_json: lifecycleJson },
         annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
     }, async (args) => guardedJson(() => applyRebuildForMcp(args.task_json, args.response_json, args.profile_json, args.capability_json, loadApprovalContext()), 'Rebuild application failed.'));

package/dist/pipeline.js

@@ -29,10 +29,16 @@ function formatFindings(findings) {
 export function isBlockingFinding(finding) {
     return finding.engine === 'ai_editor' ? finding.appliedPolicy === 'blocking' : finding.severity === 'red';
 }
-export function deriveEditScope(result) {
+export function isStrictFinding(finding) {
+    return finding.engine === 'ai_editor' || isBlockingFinding(finding);
+}
+export function strictFindings(result) {
+    return [...result.voiceDna.findings, ...result.aiEditor.findings, ...(result.editorial?.findings ?? [])].filter(isStrictFinding);
+}
+export function deriveEditScope(result, strict = false) {
     const findings = [...result.voiceDna.findings, ...result.aiEditor.findings, ...(result.editorial?.findings ?? [])];
-    const blocking = findings.filter(isBlockingFinding);
-    const pendingJudgment = findings.filter((finding) => finding.appliedPolicy === 'judgment-required');
+    const blocking = findings.filter(strict ? isStrictFinding : isBlockingFinding);
+    const pendingJudgment = strict ? [] : findings.filter((finding) => finding.appliedPolicy === 'judgment-required');
     return {
         eligibleSentenceIds: [...new Set(blocking.map((finding) => finding.sentence))].sort((left, right) => left - right),
         blocking,
@@ -41,16 +47,16 @@ export function deriveEditScope(result) {
 }
 export function renderRewritePrompt(draft, profile, result, learning = [], brief, examples = []) {
     const allFindings = [...result.voiceDna.findings, ...result.aiEditor.findings, ...(result.editorial?.findings ?? [])];
-    const scope = deriveEditScope(result);
+    const scope = deriveEditScope(result, true);
     const redFindings = scope.blocking;
-    const yellowFindings = allFindings.filter((finding) => !isBlockingFinding(finding) && finding.appliedPolicy !== 'judgment-required');
+    const yellowFindings = allFindings.filter((finding) => !isStrictFinding(finding) && finding.appliedPolicy !== 'judgment-required');
     const metrics = profile.metrics;
     return [
         '# Tier 0 — non-negotiable preservation',
         'Preserve facts, names, numbers, claims, and every unflagged sentence exactly. Do not add claims, examples, sections, hooks, or CTAs.',
         '',
-        '# Tier 1 — release blockers',
-        'Treat each blocker as a required repair. Replace the flagged sentence with a stronger, source-faithful sentence; do not merely swap one stock phrase for another.',
+        '# Tier 1 — strict repair requirements',
+        'Every active AI Editor finding is a required repair. Replace each flagged sentence with a stronger, source-faithful sentence; do not merely swap one stock phrase for another.',
         'Use only facts already present in the draft, CopySpec, WritingBrief, or supplied source context. Do not invent a source, metric, date, quotation, mechanism, example, CTA, or opinion.',
         `VoiceDNA: ${result.voiceDna.score}/100 (${result.voiceDna.passed ? 'pass' : 'fail'}).`,
         `AI Editor: ${result.aiEditor.score}/100 (${result.aiEditor.passed ? 'pass' : 'fail'}).`,
@@ -120,16 +126,18 @@ export function verify(original, candidate, profile, brief) {
     const logicLint = lintLogic(candidate, brief);
     const factLint = brief?.factSources?.length ? lintFacts({ sources: brief.factSources, draft: candidate, metadata: brief.factMetadata }) : undefined;
     const requiredFacts = verifyRequiredFacts(candidate, brief);
+    const unresolvedStrictFindings = strictFindings(checked);
     return {
         version: '2',
         original: baseline,
         candidate: checked,
         preservationScore: preservation,
         regressions,
+        strictFindings: unresolvedStrictFindings,
         finalOutput,
         logicLint,
         ...(factLint ? { factLint } : {}), ...(requiredFacts ? { requiredFacts } : {}),
-        passed: checked.passed && !regressions.some(isBlockingFinding) && preservation >= 70 && finalOutput.accepted && logicLint.passed && !factLint?.findings.some((finding) => finding.severity === 'error') && (requiredFacts?.passed ?? true),
+        passed: checked.passed && unresolvedStrictFindings.length === 0 && !regressions.some(isBlockingFinding) && preservation >= 70 && finalOutput.accepted && logicLint.passed && !factLint?.findings.some((finding) => finding.severity === 'error') && (requiredFacts?.passed ?? true),
     };
 }
 export function verifyWithCopySpec(original, candidate, profile, spec, brief) {
@@ -144,17 +152,19 @@ export function verifyRebuildWithCopySpec(original, candidate, profile, spec, br
     const logicLint = lintLogic(candidate, brief);
     const factLint = brief?.factSources?.length ? lintFacts({ sources: brief.factSources, draft: candidate, metadata: brief.factMetadata }) : undefined;
     const requiredFacts = verifyRequiredFacts(candidate, brief);
+    const unresolvedStrictFindings = strictFindings(checked);
     return {
         version: '2',
         original: baseline,
         candidate: checked,
         preservationScore: preservation,
         regressions,
+        strictFindings: unresolvedStrictFindings,
         claims,
         finalOutput: finalCheck,
         logicLint,
         ...(factLint ? { factLint } : {}), ...(requiredFacts ? { requiredFacts } : {}),
-        passed: checked.passed && !regressions.some(isBlockingFinding) && claims.passed && finalCheck.accepted && logicLint.passed && !factLint?.findings.some((finding) => finding.severity === 'error') && (requiredFacts?.passed ?? true),
+        passed: checked.passed && unresolvedStrictFindings.length === 0 && !regressions.some(isBlockingFinding) && claims.passed && finalCheck.accepted && logicLint.passed && !factLint?.findings.some((finding) => finding.severity === 'error') && (requiredFacts?.passed ?? true),
     };
 }
 function projectDeterministicVerificationArtifact(source, candidate, profile, verification, copySpec, writingBrief, verificationKind = 'claims' in verification ? 'copy_spec' : 'standard') {

package/dist/pipeline.test.js

@@ -83,7 +83,7 @@ test('orders rewrite instructions by importance tier', () => {
 });
 test('gives strong, source-faithful repair feedback without authorizing a broader rewrite', () => {
     const prompt = rewritePrompt('The scheduler failed — twice. Experts say this changes everything. The launch is on 14 August.', profile);
-    assert.match(prompt, /Treat each blocker as a required repair/);
+    assert.match(prompt, /Every active AI Editor finding is a required repair/);
     assert.match(prompt, /Do not invent a source, metric, date, quotation, mechanism, example, CTA, or opinion/);
     assert.match(prompt, /otherwise remove the unsupported framing without widening the claim/);
     assert.match(prompt, /Before responding, check every Tier 1 finding against its replacement/);
@@ -119,9 +119,11 @@ test('advisory and pending-judgment findings pass while blocking findings fail',
     const blocking = analyze('The scheduler failed — twice.', profile);
     assert.equal(blocking.passed, false);
 });
-test('verify rejects only new policy-blocking regressions', () => {
+test('verify rejects unresolved active AI Editor findings under the default strict policy', () => {
     const neutralProfile = buildProfile(['I write plainly.', 'I name the mechanism.']);
-    assert.equal(verify('The scheduler failed twice.', 'The scheduler failed twice. We leverage logs.', neutralProfile).passed, true);
+    const unresolved = verify('The scheduler failed twice.', 'The scheduler failed twice. We leverage logs.', neutralProfile);
+    assert.equal(unresolved.passed, false);
+    assert.ok(unresolved.strictFindings.some((finding) => finding.id === 'ai.leverage'));
     assert.equal(verify('The scheduler failed twice.', 'The scheduler failed — twice.', profile).passed, false);
 });
 test('keeps verify pass/fail on the legacy metric while calibration reports both metrics', () => {

package/dist/rebuild-task.test.js

@@ -174,7 +174,7 @@ test('CLI and MCP rebuild helpers share fingerprints', () => {
         version: '1', audience: 'operators', intent: 'explain', format: 'outreach',
     });
     assert.match(briefTask.prompt, /# WritingBrief/);
-    assert.equal(HYV_VERSION, '3.5.1');
+    assert.equal(HYV_VERSION, '3.6.0');
 });
 test('apply rejects forged tasks, missing capability, and substituted profiles', () => {
     const reduction = rebuildRecommendation();

package/dist/rewrite-task.js

@@ -106,7 +106,7 @@ export function prepareRewriteTask(draft, profile, copySpec, writingBrief, autho
     const result = analyze(draft, profile, writingBrief);
     const prompt = renderRewritePrompt(draft, profile, result, [], writingBrief);
     const mapped = sentences(draft);
-    const eligibleSentenceIds = new Set([...deriveEditScope(result).eligibleSentenceIds, ...authorizedSentenceIds]);
+    const eligibleSentenceIds = new Set([...deriveEditScope(result, true).eligibleSentenceIds, ...authorizedSentenceIds]);
     const taskBase = {
         version: '1',
         draft,

package/dist/rewrite-task.test.js

@@ -21,18 +21,19 @@ test('applies only eligible numbered replacements and preserves all clean bytes'
     assert.equal(result.candidate, 'I use the answer. The launch is on 14 August.');
     assert.deepEqual(result.receipt.adapterIds, []);
 });
-test('does not make an advisory broad catalog match eligible', () => {
+test('makes every active AI Editor finding eligible under the default strict rewrite policy', () => {
     const task = prepareRewriteTask('I write clear notes. This work is meaningful. I keep the mechanism visible.', profile);
-    assert.deepEqual(task.eligibleSentenceIds, []);
-    assert.deepEqual(task.sentences.map((sentence) => sentence.eligible), [false, false, false]);
+    assert.deepEqual(task.eligibleSentenceIds, [2]);
+    assert.deepEqual(task.sentences.map((sentence) => sentence.eligible), [false, true, false]);
+    assert.match(task.prompt, /Every active AI Editor finding is a required repair/);
 });
-test('grants edit scope only to blocking findings and keeps pending judgment separate', () => {
+test('makes advisory and judgment-required findings required repairs', () => {
     const advisory = prepareRewriteTask('Firstly, check the invoice.', profile);
-    assert.deepEqual(advisory.eligibleSentenceIds, []);
+    assert.deepEqual(advisory.eligibleSentenceIds, [1]);
     const neutralProfile = buildProfile(['I write plainly.', 'I name the mechanism.']);
     const pending = prepareRewriteTask('We leverage the scheduler.', neutralProfile);
-    assert.deepEqual(pending.eligibleSentenceIds, []);
-    assert.match(pending.prompt, /pending judgment/i);
+    assert.deepEqual(pending.eligibleSentenceIds, [1]);
+    assert.match(pending.prompt, /Every active AI Editor finding is a required repair/);
     const blocking = prepareRewriteTask('The scheduler failed — twice. The owner checked it.', profile);
     assert.deepEqual(blocking.eligibleSentenceIds, [1]);
 });
@@ -40,9 +41,9 @@ test('keeps blocking VoiceDNA avoid findings eligible', () => {
     const task = prepareRewriteTask('The plan uses leverage.', profile);
     assert.deepEqual(task.eligibleSentenceIds, [1]);
 });
-test('preserves founder reframes that reconciled policy makes advisory', () => {
+test('makes reconciled active founder reframes eligible under the default strict policy', () => {
     const task = prepareRewriteTask("This isn't positioning. This is proof.", profile);
-    assert.deepEqual(task.eligibleSentenceIds, []);
+    assert.deepEqual(task.eligibleSentenceIds, [1]);
 });
 test('preserves a clean draft byte-for-byte when the rewrite response is empty', () => {
     const draft = 'I write clear notes.\n\nI keep the mechanism visible.\n';

package/dist/strict-quality.js

@@ -8,8 +8,7 @@ function finding(id, disposition, reason, suggestion, sentence) {
 }
 function strictFinding(source, profile) {
     if (source.engine === 'ai_editor') {
-        const disposition = source.appliedPolicy === 'blocking' ? 'block' : 'review';
-        return finding(`strict.${source.engine}.${source.id}`, disposition, source.reason, source.suggestion, source.sentence);
+        return finding(`strict.${source.engine}.${source.id}`, 'block', source.reason, source.suggestion, source.sentence);
     }
     if (source.engine === 'voice_dna') {
         if (source.severity === 'red')

package/dist/strict-quality.test.js

@@ -45,10 +45,10 @@ test('blocks an underpowered or uncalibrated voice profile before trusting voice
     const mixedFormatReport = evaluateStrictQuality('The launch starts Tuesday.', profile(), mixedFormats);
     assert.ok(mixedFormatReport.findings.some((item) => item.id === 'strict.profile.format-spread'));
 });
-test('requires human review for a judgment-required AI pattern and blocks a configured one', () => {
+test('blocks every active AI Editor pattern under strict quality', () => {
     const review = evaluateStrictQuality('We leverage the existing checklist.', profile(), samples());
-    assert.equal(review.disposition, 'needs-human-review');
-    assert.ok(review.findings.some((item) => item.id === 'strict.ai_editor.ai.leverage' && item.disposition === 'review'));
+    assert.equal(review.disposition, 'blocked');
+    assert.ok(review.findings.some((item) => item.id === 'strict.ai_editor.ai.leverage' && item.disposition === 'block'));
     const blockedProfile = profile();
     blockedProfile.rulePolicy['ai.leverage'] = 'blocking';
     const blocked = evaluateStrictQuality('We leverage the existing checklist.', blockedProfile, samples());

package/dist/version.js

@@ -1 +1 @@
-export const HYV_VERSION = '3.5.1';
+export const HYV_VERSION = '3.6.0';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@holdyourvoice/hyv",
-  "version": "3.5.1",
+  "version": "3.6.0",
   "description": "A local-first dual-engine writing gate that protects voice and catches generic AI patterns.",
   "type": "module",
   "bin": {

package/skills/hyv-analyze/SKILL.md

@@ -23,4 +23,4 @@ hyv analyze draft.md profile.json [writing-brief.json]
 
 ## Handoff
 
-Run `hyv analyze` directly to execute the operation. For ordinary editing, follow on with hyv-rewrite-prompt or hyv-prepare-rewrite. When the caller asks for a strict AI-pattern and voice-match decision, hand off to hyv-strict-check with a calibrated Profile v3 and the required local samples; analysis alone is not a strict-ready result.
+Run `hyv analyze` directly to execute the operation. Follow with hyv-rewrite-prompt or hyv-prepare-rewrite: both use strict-by-default repair scope, so every active AI Editor finding becomes a required repair. For calibrated voice evidence, hand off to hyv-strict-check with a Profile v3 and the required local samples; analysis alone is not a strict-ready result.

package/skills/hyv-analyze/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Analyze"
   short_description: "Run separate VoiceDNA and AI Editor checks plus a non-scoring Unicode hygiene inspection against a draft."
-  default_prompt: "Use $hyv-analyze to run separate voicedna and ai editor checks plus a non-scoring unicode hygiene inspection against a draft."
+  default_prompt: "Use $hyv-analyze to inspect this draft, then use the strict-by-default rewrite workflow so every active AI Editor finding is repaired."

package/skills/hyv-apply-rebuild/SKILL.md

@@ -16,6 +16,7 @@ hyv apply-rebuild task.json response.json profile.json (--capability-stdin|--cap
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: evaluation rejects a rebuild candidate with any unresolved active AI Editor finding.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-apply-rebuild/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Apply Rebuild"
   short_description: "Validate and evaluate a whole-document rebuild response against a prepared authorized rebuild task."
-  default_prompt: "Use $hyv-apply-rebuild to validate and evaluate a whole-document rebuild response against a prepared authorized rebuild task."
+  default_prompt: "Use $hyv-apply-rebuild to reject a whole-document rebuild response with any unresolved active AI Editor finding."

package/skills/hyv-apply-rewrite/SKILL.md

@@ -16,6 +16,7 @@ hyv apply-rewrite task.json response.json profile.json
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- The task's default strict policy rejects the candidate at deterministic verification when an active AI Editor finding remains. A successful deterministic result is not calibrated `strict-ready` voice evidence.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-apply-rewrite/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Apply Rewrite"
   short_description: "Validate and apply a model response to a prepared rewrite task, then run local gates."
-  default_prompt: "Use $hyv-apply-rewrite to validate and apply a model response to a prepared rewrite task, then run local gates."
+  default_prompt: "Use $hyv-apply-rewrite to validate a response through the strict default local gates."

package/skills/hyv-prepare-rebuild/SKILL.md

@@ -16,6 +16,7 @@ hyv prepare-rebuild draft.md profile.json reduction.json copy-spec.json task.jso
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: the later rebuild evaluation rejects a candidate with any unresolved active AI Editor finding.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-prepare-rebuild/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Prepare Rebuild"
   short_description: "Prepare a rebuild task only after an upstream REBUILD recommendation, a CopySpec, and a signed rebuild-authorization capability."
-  default_prompt: "Use $hyv-prepare-rebuild to prepare a rebuild task only after an upstream rebuild recommendation, a copyspec, and a signed rebuild-authorization capability."
+  default_prompt: "Use $hyv-prepare-rebuild to prepare a strict rebuild task only after an upstream rebuild recommendation, a copyspec, and a signed rebuild-authorization capability."

package/skills/hyv-prepare-rewrite/SKILL.md

@@ -16,6 +16,7 @@ hyv prepare-rewrite draft.md profile.json task.json [copy-spec.json] [writing-br
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: the task authorizes repairs for every active AI Editor finding and keeps unrelated sentences locked.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-prepare-rewrite/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Prepare Rewrite"
   short_description: "Prepare a local, versioned, fingerprint-bound rewrite task."
-  default_prompt: "Use $hyv-prepare-rewrite to prepare a local, versioned, fingerprint-bound rewrite task."
+  default_prompt: "Use $hyv-prepare-rewrite to prepare a strict local, versioned, fingerprint-bound rewrite task."

package/skills/hyv-rebuild-writer-request/SKILL.md

@@ -16,6 +16,7 @@ hyv rebuild-writer-request task.json writer-request.json
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- The resulting whole-document candidate remains subject to strict default verification: every active AI Editor finding must be repaired before it can proceed.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-rebuild-writer-request/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Rebuild Writer Request"
   short_description: "Create the writer-only payload for a prepared rebuild task."
-  default_prompt: "Use $hyv-rebuild-writer-request to create the writer-only payload for a prepared rebuild task."
+  default_prompt: "Use $hyv-rebuild-writer-request to create the writer-only payload for a prepared rebuild task that will receive strict default verification."

package/skills/hyv-rewrite-prompt/SKILL.md

@@ -1,11 +1,11 @@
 ---
 name: hyv-rewrite-prompt
-description: Create a constrained editing brief without rewriting the draft.
+description: Create a strict constrained editing brief without rewriting the draft.
 ---
 
 # hyv-rewrite-prompt
 
-Create a constrained editing brief. It does not rewrite the draft or call a model. The brief is printed to stdout for redirect into a markdown file.
+Create a strict constrained editing brief. It does not rewrite the draft or call a model. The brief is printed to stdout for redirect into a markdown file.
 
 ## Usage
 
@@ -16,9 +16,10 @@ hyv rewrite-prompt draft.md profile.json [writing-brief.json] > rewrite-brief.md
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: every active AI Editor finding is a required, source-faithful repair. Do not return a candidate with an active finding unresolved; a separate calibrated `hyv-strict-check` is required only when the caller needs a `strict-ready` voice-evidence result.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 
 ## Handoff
 
-Run `hyv rewrite-prompt` directly to execute the operation. Follow-on agents: hyv-prepare-rewrite.
+Run `hyv rewrite-prompt` directly to execute the operation. Follow-on agents: hyv-prepare-rewrite, hyv-verify.

package/skills/hyv-rewrite-prompt/agent.json

@@ -2,7 +2,7 @@
   "schema_version": "1.0.0",
   "id": "hyv-rewrite-prompt",
   "title": "hyv-rewrite-prompt",
-  "description": "Create a constrained editing brief without rewriting the draft.",
+  "description": "Create a strict constrained editing brief without rewriting the draft.",
   "instruction_file": "SKILL.md",
   "role": "Brief author",
   "workflow_phase": "plan",

package/skills/hyv-rewrite-prompt/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Rewrite Prompt"
-  short_description: "Create a constrained editing brief without rewriting the draft."
-  default_prompt: "Use $hyv-rewrite-prompt to create a constrained editing brief without rewriting the draft."
+  short_description: "Create a strict editing brief without rewriting the draft."
+  default_prompt: "Use $hyv-rewrite-prompt to create a strict constrained editing brief that requires repair of every active AI Editor finding."

package/skills/hyv-strict-check/SKILL.md

@@ -5,7 +5,7 @@ description: Run the local strict AI-pattern and calibrated voice-match gate.
 
 # hyv-strict-check
 
-Run the opt-in strict local quality gate without changing the draft, profile, or learning state.
+Run the calibrated strict local quality gate without changing the draft, profile, or learning state.
 
 ## Usage
 
@@ -17,8 +17,8 @@ hyv strict-check draft.md profile-v3.json sample-a.md sample-b.md sample-c.md sa
 
 - Requires a Profile v3 built from at least five samples, five or more local validation samples, 1,500 validation-sample words, no duplicate samples, a consistent visible format baseline, and calibrated fingerprint tolerances.
 - Returns `strict-ready`, `needs-human-review`, or `blocked` as JSON and exits `0` only for `strict-ready`.
-- `blocked` covers missing strict evidence, configured blocking AI patterns, hard calibrated fingerprint drift, red VoiceDNA findings, and unresolved hidden text.
-- `needs-human-review` keeps advisory and judgment-required AI patterns visible. It is not a clean result.
+- `blocked` covers missing strict evidence, every active AI Editor pattern, hard calibrated fingerprint drift, red VoiceDNA findings, and unresolved hidden text.
+- `needs-human-review` covers remaining non-blocking VoiceDNA or editorial review findings. It is not a clean result.
 - It remains local-first and never sends samples or drafts to a provider.
 
 ## Handoff

package/skills/hyv-verify/SKILL.md

@@ -16,9 +16,10 @@ hyv verify original.md candidate.md profile.json [writing-brief.json]
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: verification fails when the candidate has any active AI Editor finding, including an advisory or judgment-required match. An explicit Profile v3 `disabled` policy remains a deliberate user exception.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 
 ## Handoff
 
-Run `hyv verify` directly to execute the operation. If the candidate must satisfy the strict AI-pattern and voice-match policy, run hyv-strict-check before any lifecycle handoff. A passing verify report does not substitute for strict-ready. Follow-on agents: hyv-strict-check, hyv-lifecycle.
+Run `hyv verify` directly to execute the operation. For calibrated voice evidence, run hyv-strict-check before any lifecycle handoff. A passing verify report does not substitute for strict-ready. Follow-on agents: hyv-strict-check, hyv-lifecycle.

package/skills/hyv-verify/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Verify"
   short_description: "Verify a revised candidate against an original draft and portable profile."
-  default_prompt: "Use $hyv-verify to verify a revised candidate against an original draft and portable profile."
+  default_prompt: "Use $hyv-verify to reject any revised candidate with an unresolved active AI Editor finding."

package/skills/hyv-verify-spec/SKILL.md

@@ -1,11 +1,11 @@
 ---
 name: hyv-verify-spec
-description: Verify a candidate against the existing voice gates and a local CopySpec.
+description: Verify a candidate against strict default voice gates and a local CopySpec.
 ---
 
 # hyv-verify-spec
 
-Verify a candidate against the existing voice gates and a local CopySpec. Immutable claims remain verbatim unless atoms are supplied; then each declared atom must remain. Prohibited claims fail closed.
+Verify a candidate against strict default voice gates and a local CopySpec. Immutable claims remain verbatim unless atoms are supplied; then each declared atom must remain. Prohibited claims fail closed.
 
 ## Usage
 
@@ -16,6 +16,7 @@ hyv verify-spec original.md candidate.md profile.json copy-spec.json [writing-br
 ## Behavior
 
 - Deterministic and local-first: this command never calls a provider and never sends drafts, samples, profiles, or telemetry to a service.
+- Strict by default: verification fails when the candidate has any active AI Editor finding, in addition to CopySpec failures.
 - Commands that write use an explicit output path. Confirm that path before running the command.
 - This agent describes how to invoke the command and what it returns. It does not change command behavior or exit codes.
 

package/skills/hyv-verify-spec/agent.json

@@ -2,7 +2,7 @@
   "schema_version": "1.0.0",
   "id": "hyv-verify-spec",
   "title": "hyv-verify-spec",
-  "description": "Verify a candidate against the existing voice gates and a local CopySpec.",
+  "description": "Verify a candidate against strict default voice gates and a local CopySpec.",
   "instruction_file": "SKILL.md",
   "role": "CopySpec verifier",
   "workflow_phase": "verify",

package/skills/hyv-verify-spec/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "Hyv Verify Spec"
-  short_description: "Verify a candidate against the existing voice gates and a local CopySpec."
-  default_prompt: "Use $hyv-verify-spec to verify a candidate against the existing voice gates and a local copyspec."
+  short_description: "Verify a candidate against strict default voice gates and a local CopySpec."
+  default_prompt: "Use $hyv-verify-spec to reject a candidate with any active AI Editor finding or CopySpec failure."