harness-everything 0.41.3 → 0.42.0

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
       "name": "harness-everything",
       "source": "./",
       "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-      "version": "0.41.3",
+      "version": "0.42.0",
       "author": {
         "name": "Miya Daniel"
       },

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "harness-everything",
   "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-  "version": "0.41.3",
+  "version": "0.42.0",
   "author": {
     "name": "Miya Daniel"
   },

package/CHANGELOG.md

@@ -2,6 +2,12 @@
 
 All notable changes to this project are documented in this file.
 
+## [0.42.0](https://github.com/dyphn1/Harness-everything/compare/v0.41.3...v0.42.0) (2026-10-25)
+
+### Features
+
+* **opencode:** wire bounded reconciliation ([039cf21](https://github.com/dyphn1/Harness-everything/commit/039cf2194ea232dc5bea715b67684a234a6a1ac4)), closes [#341](https://github.com/dyphn1/Harness-everything/issues/341) [#325](https://github.com/dyphn1/Harness-everything/issues/325)
+
 ## [0.41.3](https://github.com/dyphn1/Harness-everything/compare/v0.41.2...v0.41.3) (2026-10-25)
 
 ### Bug Fixes

package/benchmarks/results/h1-provider-independence-qualification/README.md

@@ -0,0 +1,72 @@
+# H1 provider-independence qualification (preparation)
+
+This directory records the bounded #325 H1 preparation/non-interference qualification scope. It is preparation evidence only; it is not full H1, Phase 3c, E1, paired evidence, benefit evidence, or #320 promotion.
+
+## Scope
+
+- H1 preparation/non-interference only.
+- `suggestDocProviders` is called as an independent side input. Suggestions and provider identity are never passed to the F4 consumer, workflow, binding, or reconciliation path.
+- No runtime, shared harness, package, manifest, lockfile, fixture JSON, or issue edits are part of this qualification.
+- The registered F4 evidence and executor remain the only evidence path: executor `f4-pos-body-edit-session`; provenance `claude-code-capture:<captureId>`.
+
+## Matrix
+
+Cross product: 3 provider inputs × 2 evidence modes × 2 distributions = 12 cells.
+
+Provider inputs:
+
+- A: only `fixture-lint-a` metadata; expected exactly the A suggestion.
+- B: only `fixture-lint-b` metadata; expected exactly the B suggestion.
+- Omitted: no provider metadata; expected `suggestions: []` and one unavailable `doc-lint` record.
+
+Evidence modes:
+
+- Valid evidence: actual registered doc-check `hostRun` and correlated `PostToolUse` capture.
+- Zero evidence: same F4-POS-body-edit task definition and loaded `verification-loop` binding, with only `lint-run` and `post-lint` removed.
+
+Distributions: `source` and `packaged`.
+
+## Fixed expected full applicable MUST objects
+
+- `{id:"doc-change:doc-lint", capability:"doc-lint", strength:"MUST", applicability:"applicable", scope:{paths:["docs/a.md"]}, acceptance:"Run document lifecycle lint for every changed governed document."}`
+- `{id:"doc-change:link-check", capability:"link-check", strength:"MUST", applicability:"applicable", scope:{paths:["docs/a.md"]}, acceptance:"Check in-repository links for every changed governed document."}`
+
+`expectedMustIds` must equal those IDs. The pilot runtime report `results` field is empty and is not asserted. A non-blocking missing-path-coverage diagnostic for the SHOULD `prose-reminder` is expected; diagnostics are not asserted empty.
+
+## Asserted fields
+
+Valid evidence cells assert:
+
+- `reportOrigin: runtime`
+- `expectedMustIds` exact
+- no unsatisfied MUSTs
+- `creditedCapabilities` and accepted refs for both capabilities
+- `pilotCredit: true`, `pilotStatus: credited`
+- verification step pass
+- Stop exit `0`, workflow state `satisfied`
+
+Zero evidence cells assert:
+
+- both MUSTs remain expected and unsatisfied
+- `creditedCapabilities` and `acceptedRefs` empty
+- `pilotCredit: false`, `pilotStatus: unresolved`
+- verification step pending
+- Stop exit `0`, workflow state `running` (not satisfied)
+
+## Test approach and execution
+
+This is an integration qualification because each F4 cell crosses real child processes, filesystem state, hook entries, and workflow disposition. Expectations come from #325 H1 and the fixed F4 fixture. The existing valid-evidence and zero-evidence cases are positive and negative controls; this does not manufacture a runtime-bug RED or authorize runtime edits.
+
+| Command | Result |
+|---|---|
+| `node ci/mechanism-325-3b-h1-qualification.test.js` | 36 PASS, 0 FAIL, 0 INFRA; all 12 matrix cells and source/package MUST derivations passed |
+| Initial `npm run verify:all -- --fast` | 3 PASS, 2 FAIL, 11 UNCHECKED. Environment preflight failure: this worktree had no node_modules; self-regression lacked node_modules and the #341 C2 AST checker/mutation controls reported AST_PARSER_UNAVAILABLE. This is not counted as a product defect. |
+| Focused dependency controls after installation | Node 24.22.1: mechanism-2i-crlf-frontmatter passed; mechanism-341-boundary-enforcement passed all C1/C2 controls; npm run test:mutations passed 113 mutants and 2 source probe suites. |
+| Final `npm run verify:all` | Node 24.22.1: 13 PASS, 0 FAIL, 3 UNCHECKED. The unchecked items are Node 22 self-regression smoke, Node 22 installer roundtrip, and npm ci under Node 22. No CI or live-host evidence is claimed. |
+
+The initial fast run was retained as the environment-preflight result and was not rerun. After confirming the missing dependencies with npm ls and resolving pinned acorn/acorn-walk, the full verification gate passed. Dependencies were installed with npm ci --ignore-scripts; the only package lifecycle script is prepare: husky install.
+
+## Pending
+
+- Full H1 and Phase 3c remain pending.
+- E1, paired evidence, benefit evidence, and #320 promotion remain pending.

package/ci/mechanism-325-3b-h1-qualification.test.js

@@ -0,0 +1,206 @@
+#!/usr/bin/env node
+'use strict';
+
+// #325 H1 deterministic qualification: provider independence/preparation.
+// H1 only: advisory suggestions are never passed to the F4 consumer.
+
+const fs = require('fs');
+const path = require('path');
+
+const fx = require(path.join(__dirname, 'lib', 'doc-reconciliation-prep-fixtures'));
+const { stage } = require(path.join(__dirname, 'lib', 'doc-reconciliation-staged-tree'));
+const F4_CASES = path.join(__dirname, 'fixtures', 'doc-obligations', 'reconciliation-s2b', 'f4', 'cases.json');
+
+const REPO_ROOT = path.resolve(__dirname, '..');
+const SOURCE_OBLIGATIONS = require(path.join(REPO_ROOT, 'harness-everything', 'scripts', 'doc-obligations.js'));
+const PACKAGED_OBLIGATIONS = require(path.join(REPO_ROOT, 'plugins', 'harness-everything', 'skills', 'harness-everything', 'scripts', 'doc-obligations.js'));
+const SOURCE_PROVIDER = require(path.join(REPO_ROOT, 'harness-everything', 'scripts', 'doc-providers.js'));
+const PACKAGED_PROVIDER = require(path.join(REPO_ROOT, 'plugins', 'harness-everything', 'skills', 'harness-everything', 'scripts', 'doc-providers.js'));
+
+const LEGS = ['source', 'packaged'];
+const CASE = id => fx.readCaseCorpus(F4_CASES).cases.find(item => item.id === id);
+
+const EXPECTED_MUSTS = [
+  { id: 'doc-change:doc-lint', capability: 'doc-lint', strength: 'MUST', applicability: 'applicable', scope: { paths: ['docs/a.md'] }, acceptance: 'Run document lifecycle lint for every changed governed document.' },
+  { id: 'doc-change:link-check', capability: 'link-check', strength: 'MUST', applicability: 'applicable', scope: { paths: ['docs/a.md'] }, acceptance: 'Check in-repository links for every changed governed document.' }
+];
+
+const PROVIDER_A_INPUT = {
+  activeStep: { id: 'verify-docs', stepType: 'verification' },
+  neededCapabilities: ['doc-lint'],
+  availableProviderMetadata: [
+    { providerId: 'fixture-lint-a', capabilities: ['doc-lint'], stepTypes: ['verification'], requiresHostCapabilities: ['execution-events'], locator: { kind: 'tool', id: 'fixture-lint-a' } }
+  ],
+  hostCapabilities: ['execution-events']
+};
+
+const PROVIDER_B_INPUT = {
+  activeStep: { id: 'verify-docs', stepType: 'verification' },
+  neededCapabilities: ['doc-lint'],
+  availableProviderMetadata: [
+    { providerId: 'fixture-lint-b', capabilities: ['doc-lint'], stepTypes: ['verification'], requiresHostCapabilities: ['execution-events'], locator: { kind: 'tool', id: 'fixture-lint-b' } }
+  ],
+  hostCapabilities: ['execution-events']
+};
+
+const PROVIDER_OMITTED_INPUT = {
+  activeStep: { id: 'verify-docs', stepType: 'verification' },
+  neededCapabilities: ['doc-lint'],
+  availableProviderMetadata: [],
+  hostCapabilities: []
+};
+
+const PROVIDER_A_SUGGESTION = [{ capability: 'doc-lint', providerId: 'fixture-lint-a', availability: 'available', reason: 'active-step capability and host requirements match' }];
+const PROVIDER_B_SUGGESTION = [{ capability: 'doc-lint', providerId: 'fixture-lint-b', availability: 'available', reason: 'active-step capability and host requirements match' }];
+const PROVIDER_OMITTED_UNAVAILABLE = [{ capability: 'doc-lint', reason: 'no available provider metadata for the active-step capability' }];
+
+let passCount = 0;
+let failCount = 0;
+let infraCount = 0;
+
+function deepEqual(left, right) {
+  return JSON.stringify(left) === JSON.stringify(right);
+}
+
+function record(status, id, detail) {
+  const suffix = detail ? ` - ${detail}` : '';
+  if (status === 'PASS') passCount += 1;
+  else if (status === 'FAIL') failCount += 1;
+  else infraCount += 1;
+  console.log(`   ${status.padEnd(5)} ${id}${suffix}`);
+}
+
+function cloneZero(caseDef) {
+  return {
+    ...caseDef,
+    id: `${caseDef.id}-zero-evidence`,
+    actions: caseDef.actions.filter(action => action.id !== 'lint-run' && action.id !== 'post-lint')
+  };
+}
+
+function checkProvider(leg, providerCase) {
+  const module = leg === 'source' ? SOURCE_PROVIDER : PACKAGED_PROVIDER;
+  let input;
+  let expectedSuggestions;
+  let expectedUnavailable;
+  if (providerCase === 'A') {
+    input = PROVIDER_A_INPUT;
+    expectedSuggestions = PROVIDER_A_SUGGESTION;
+    expectedUnavailable = [];
+  } else if (providerCase === 'B') {
+    input = PROVIDER_B_INPUT;
+    expectedSuggestions = PROVIDER_B_SUGGESTION;
+    expectedUnavailable = [];
+  } else {
+    input = PROVIDER_OMITTED_INPUT;
+    expectedSuggestions = [];
+    expectedUnavailable = PROVIDER_OMITTED_UNAVAILABLE;
+  }
+  const result = module.suggestDocProviders(input);
+  const ok = deepEqual(result.suggestions, expectedSuggestions) && deepEqual(result.unavailable, expectedUnavailable);
+  record(ok ? 'PASS' : 'FAIL', `H1.${leg}.${providerCase}.provider-suggestion`, ok ? '' : `suggestions=${JSON.stringify(result.suggestions)}, unavailable=${JSON.stringify(result.unavailable)}`);
+  return ok;
+}
+
+function checkDerivedMusts(leg, label) {
+  const module = leg === 'source' ? SOURCE_OBLIGATIONS : PACKAGED_OBLIGATIONS;
+  const caseDef = CASE('F4-POS-body-edit');
+  const config = JSON.parse(caseDef.input.baseFiles['.agents/harness-everything/project-docs.json']);
+  const base = { files: caseDef.input.baseFiles };
+  const candidate = { files: { ...caseDef.input.baseFiles, ...caseDef.input.candidateFiles } };
+  const { facts } = module.deriveDocFacts({ base, candidate, config, inventory: { complete: true } });
+  const obligations = module.deriveDocObligations(facts);
+  const fullMusts = obligations.filter(item => item.strength === 'MUST' && item.applicability === 'applicable');
+
+  let ok = deepEqual(fullMusts, EXPECTED_MUSTS);
+  for (const item of fullMusts) {
+    if (!item.id || !item.capability || !item.strength || !item.applicability || !item.scope || !Array.isArray(item.scope.paths) || item.scope.paths.some(pathValue => !pathValue) || !item.acceptance) ok = false;
+  }
+  record(ok ? 'PASS' : 'FAIL', `${label}.derived-full-musts`, ok ? '' : JSON.stringify(fullMusts));
+  return ok;
+}
+
+function runCell(leg, providerCase, evidenceMode) {
+  const label = `H1.${leg}.${providerCase}.${evidenceMode}`;
+  const staged = stage(leg);
+  try {
+    checkProvider(leg, providerCase);
+    checkDerivedMusts(leg, label);
+
+    staged.writeExport(staged.validExport());
+
+    const baseCase = CASE('F4-POS-body-edit');
+    const caseDef = evidenceMode === 'valid' ? baseCase : cloneZero(baseCase);
+
+    const root = fx.replayCase(caseDef, staged.dist, { pilot: '1' });
+    const faults = fx.infrastructureFaults(caseDef, root, 'on');
+    if (faults.length) {
+      record('INFRA', label, faults.map(fault => `${fault.actionId}: ${fault.message}`).join('; '));
+      return;
+    }
+
+    const pilot = root.pilot || {};
+    const checks = [];
+
+    checks.push(deepEqual(root.pilotReportOrigin, 'runtime'));
+    checks.push(deepEqual(pilot.reportOrigin, 'runtime'));
+    checks.push(deepEqual(pilot.expectedMustIds, ['doc-change:doc-lint', 'doc-change:link-check']));
+    const verifyAction = root.actions && root.actions['verify-pass'];
+    const verifyRequirement = verifyAction?.state?.obligations?.requirements?.find(item => item.id === 'req-verify');
+    checks.push(verifyRequirement?.requiredBindings?.some(binding => binding.id === 'verification-loop' && binding.status === 'loaded') === true);
+
+    if (evidenceMode === 'valid') {
+      checks.push(pilot.pilotCredit === true);
+      checks.push(pilot.pilotStatus === 'credited');
+      checks.push(deepEqual(pilot.unsatisfied, []));
+      checks.push(deepEqual(pilot.creditedCapabilities, ['doc-lint', 'link-check']));
+      checks.push(Array.isArray(pilot.acceptedRefs) && pilot.acceptedRefs.some(ref => ref.endsWith('#doc-lint')) && pilot.acceptedRefs.some(ref => ref.endsWith('#link-check')));
+      checks.push(verifyAction && verifyAction.exitCode === 0);
+      const stopAction = root.actions && root.actions.stop;
+      checks.push(stopAction && stopAction.exitCode === 0);
+      checks.push(stopAction && stopAction.state && stopAction.state.workflow && stopAction.state.workflow.state === 'satisfied');
+    } else {
+      checks.push(pilot.pilotCredit === false);
+      checks.push(pilot.pilotStatus === 'unresolved');
+      checks.push(Array.isArray(pilot.unsatisfied) && pilot.unsatisfied.length >= 2);
+      checks.push(pilot.unsatisfied.some(item => item.obligationId === 'doc-change:doc-lint'));
+      checks.push(pilot.unsatisfied.some(item => item.obligationId === 'doc-change:link-check'));
+      checks.push(deepEqual(pilot.creditedCapabilities, []));
+      checks.push(deepEqual(pilot.acceptedRefs, []));
+      checks.push(verifyAction && verifyAction.exitCode === 2);
+      const stopAction = root.actions && root.actions.stop;
+      checks.push(stopAction && stopAction.exitCode === 0);
+      checks.push(stopAction && stopAction.state && stopAction.state.workflow && stopAction.state.workflow.state === 'running');
+    }
+
+    record(checks.every(Boolean) ? 'PASS' : 'FAIL', label, checks.every(Boolean) ? '' : `checks=${JSON.stringify(checks)}, pilot=${JSON.stringify(pilot).slice(0, 500)}`);
+  } catch (error) {
+    record('FAIL', label, error.message);
+  } finally {
+    staged.cleanup();
+  }
+}
+
+function main() {
+  console.log('\n== H1 qualification: provider side input and F4 replay matrix ==');
+  for (const leg of LEGS) {
+    for (const providerCase of ['A', 'B', 'omitted']) {
+      for (const evidenceMode of ['valid', 'zero']) {
+        runCell(leg, providerCase, evidenceMode);
+      }
+    }
+  }
+
+  console.log('\n== H1 qualification evidence ==');
+  console.log(`passing checks: ${passCount}`);
+  console.log(`unexpected failures: ${failCount}`);
+  console.log(`infrastructure faults: ${infraCount}`);
+  if (failCount === 0 && infraCount === 0) {
+    console.log('\nSTATUS: GREEN (all assertions passed)');
+    return 0;
+  }
+  console.log('\nSTATUS: NOT-GREEN');
+  return 1;
+}
+
+process.exitCode = main();