harness-everything 0.12.1 → 0.12.2

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": "0.12.1",
+      "version": "0.12.2",
       "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": "0.12.1",
+  "version": "0.12.2",
   "author": {
     "name": "Miya Daniel"
   },

package/CHANGELOG.md

@@ -2,6 +2,15 @@
 
 All notable changes to this project are documented in this file.
 
+## [0.12.2](https://github.com/dyphn1/Harness-everything/compare/v0.12.1...v0.12.2) (2026-09-19)
+
+### Bug Fixes
+
+* **workflow:** admit observation-only verification at limit ([08b859e](https://github.com/dyphn1/Harness-everything/commit/08b859ef27c04a07d9a0a6faeeed52f3da986378))
+* **workflow:** admit observation-only verification at limit ([8924a3c](https://github.com/dyphn1/Harness-everything/commit/8924a3cc80455592accbe81e3f2df0f4479a2d8b))
+* **workflow:** allow verification probe at iteration limit ([7ce511e](https://github.com/dyphn1/Harness-everything/commit/7ce511e29ca368371d996f4cbeff02298d97434c))
+* **workflow:** allow verification probe at iteration limit ([07e6ed0](https://github.com/dyphn1/Harness-everything/commit/07e6ed0b369e38f274ac099af590b12373b1ad1c))
+
 ## [0.12.1](https://github.com/dyphn1/Harness-everything/compare/v0.12.0...v0.12.1) (2026-09-19)
 
 ### Bug Fixes

package/ci/mechanism-2ah-workflow-lifecycle.test.js

@@ -203,6 +203,58 @@ try {
   git(['init']); fs.writeFileSync(path.join(repo, 'README.md'), 'fixture\n'); git(['add', '.']);
   git(['-c', 'user.name=Harness Test', '-c', 'user.email=harness@example.invalid', 'commit', '-m', 'fixture']);
   git(['worktree', 'add', linked, '-b', 'isolated']);
+
+  // #164: the mutation budget is an edit budget, not a reason to make the
+  // final edit unverifiable. At the exact limit a verification-shaped shell
+  // gets observation-only admission. No-effect verification may complete the
+  // workflow; a verification command that mutates still exhausts after Post.
+  const finalVerify = fixtureSession('workflow-final-iteration-verify');
+  check(finalVerify.route('Fix this checkout bug with a regression test').status === 0 &&
+    finalVerify.control('start').status === 0,
+    '#164 final-iteration verification fixture starts Tier 2');
+  const finalVerifyLimit = read(finalVerify.file).workflowPlan.limits.maxIterations;
+  for (let i = 0; i < finalVerifyLimit; i++) {
+    check(finalVerify.gate('Write', { file_path: path.join(repo, 'final-verify-' + i + '.txt') }).status === 0,
+      '#164 confirmed mutation fills iteration budget ' + (i + 1) + '/' + finalVerifyLimit);
+  }
+  check(read(finalVerify.file).budget.counters.iterations === finalVerifyLimit,
+    '#164 fixture reaches the exact mutation limit while still running');
+  const finalVerifyId = 'toolu_final_iteration_verify';
+  check(finalVerify.gate('Bash', { command: 'node run-test.js' }, { tool_use_id: finalVerifyId }).status === 0,
+    '#164 no-effect verification shell is admitted at the exact mutation limit');
+  check(finalVerify.persist('Bash', { command: 'node run-test.js' }, { stdout: 'tests passed' },
+    { tool_use_id: finalVerifyId }).status === 0,
+    '#164 no-effect verification settles without exhausting the budget');
+  check(read(finalVerify.file).budget.counters.iterations === finalVerifyLimit &&
+    read(finalVerify.file).state === 'running',
+    '#164 no-effect verification preserves the full mutation count and running state');
+  check(finalVerify.stop().status === 0 && read(finalVerify.file).state === 'satisfied',
+    '#164 verification after the final allowed edit can satisfy Stop');
+
+  const finalMutator = fixtureSession('workflow-final-iteration-mutator');
+  check(finalMutator.route('Fix this checkout bug with a regression test').status === 0 &&
+    finalMutator.control('start').status === 0,
+    '#164 at-limit mutator fixture starts Tier 2');
+  const finalMutatorLimit = read(finalMutator.file).workflowPlan.limits.maxIterations;
+  for (let i = 0; i < finalMutatorLimit; i++) {
+    check(finalMutator.gate('Write', { file_path: path.join(repo, 'final-mutator-' + i + '.txt') }).status === 0,
+      '#164 mutator fixture fills iteration budget ' + (i + 1) + '/' + finalMutatorLimit);
+  }
+  const finalMutatorId = 'toolu_final_iteration_mutator';
+  check(finalMutator.gate('Bash', { command: 'node run-test-and-mutate.js' }, { tool_use_id: finalMutatorId }).status === 0,
+    '#164 verification-shaped command receives observation-only admission at limit');
+  const finalOverflowTarget = path.join(repo, 'final-iteration-overflow.txt');
+  fs.writeFileSync(finalOverflowTarget, 'mutation beyond limit\n');
+  check(finalMutator.persist('Bash', { command: 'node run-test-and-mutate.js' }, { stdout: 'changed' },
+    { tool_use_id: finalMutatorId }).status === 2,
+    '#164 at-limit verification-shaped command that mutates is rejected on observation');
+  const finalMutatorState = read(finalMutator.file);
+  check(finalMutatorState.state === 'blocked' &&
+    finalMutatorState.budget?.state === 'budget-exhausted' &&
+    finalMutatorState.budget?.reasonCode === 'iteration-budget-exhausted',
+    '#164 observed mutation beyond the limit leaves an auditable blocked workflow');
+  fs.unlinkSync(finalOverflowTarget);
+
   const taskNotification = '<task-notification><summary>Background command "Run full test suite" failed with exit code 1</summary></task-notification>';
   const unboundNotification = node('harness-everything/scripts/kernel-router.js', { cwd: repo, prompt: taskNotification });
   check(unboundNotification.status === 0 &&

package/hooks/scripts/lib/workflow-isolation.js

@@ -135,6 +135,11 @@ function linkedWorktree(cwd, boundRoot) {
 // Unsupported shell syntax still fails closed.
 const DANGEROUS_FLAG_RE = /\s(?:--(?:output|ext-diff|textconv|exec|pre|pre-glob|pager|open|batch|filters)|-[xoO])(?:\b|=)|\s-(?:exec|execdir|delete|fprint|fprintf)\b/i;
 const READ_ONLY_PREFIX_RE = /^(?:git\s+(?:status|rev-parse|diff|log|show|ls-files|check-ignore)(?:\s|$)|git\s+branch\s+--show-current$|git\s+worktree\s+list(?:\s|$)|gh\s+(?:issue|pr)\s+(?:view|list)(?:\s|$)|gh\s+repo\s+view(?:\s|$)|gh\s+auth\s+status(?:\s|$)|(?:pwd|ls|dir|cat|type|head|tail|wc|stat|rg|grep|echo|Get-Location|Get-ChildItem|Get-Content|Select-String|Test-Path)(?:\s|$))/i;
+const VERIFY_COMMAND_RE = /\b(test|spec|jest|vitest|mocha|pytest|rspec|phpunit|tsc|eslint|lint|build|compile|verify|check)\b/i;
+
+function isVerificationShell(command) {
+  return VERIFY_COMMAND_RE.test(String(command || ''));
+}
 
 function isReadOnlySegment(segment) {
   return READ_ONLY_PREFIX_RE.test(segment) && !DANGEROUS_FLAG_RE.test(segment);
@@ -311,4 +316,4 @@ function assertShellScope(command, cwd, isolatedRoot) {
   }
 }
 
-module.exports = { canonical, key, within, linkedWorktree, classifyShell, workspaceFingerprint, shellProbeKey, inputOf, cwdOf, commandOf, mutationPaths, worktreeRoots, directMutationInWorkspace, assertTargets, assertShellScope };
+module.exports = { canonical, key, within, linkedWorktree, classifyShell, isVerificationShell, workspaceFingerprint, shellProbeKey, inputOf, cwdOf, commandOf, mutationPaths, worktreeRoots, directMutationInWorkspace, assertTargets, assertShellScope };

package/hooks/scripts/lib/workflow-runtime.js

@@ -196,14 +196,30 @@ function reclaimStaleMutationProbes(context, keepKey) {
   }
 }
 
-function registerMutationProbe(context, key, fingerprint, reserveIteration, observationCwd, timeoutMs) {
+function registerMutationProbe(context, key, fingerprint, reserveIteration, observationCwd, options = {}) {
   if (!MUTATION_PROBE_KEY.test(String(key || '')) || !MUTATION_PROBE_KEY.test(String(fingerprint || ''))) {
     throw new Error('invalid mutation probe identity');
   }
   return withMutationProbeLock(context, () => {
     refreshWorkflowForProbe(context);
     reclaimStaleMutationProbes(context, key);
-    if (reserveIteration) assertIterationCapacity(context, 'shell:untrusted');
+    let activeReservation = Boolean(reserveIteration);
+    let atLimitAllowance = false;
+    if (activeReservation) {
+      const budget = ensureWorkflowBudget(context.workflow);
+      const limit = budget.limits.maxIterations;
+      const reserved = mutationProbeReservations(context);
+      if (options.allowAtLimit === true && budget.state === 'active' && limit !== null &&
+          budget.counters.iterations === limit && reserved === 0) {
+        // Verification-shaped commands get observation-only admission at the
+        // exact mutation limit. If they mutate, settlement still records the
+        // mutation and exhausts the budget after execution (#164).
+        activeReservation = false;
+        atLimitAllowance = true;
+      } else {
+        assertIterationCapacity(context, 'shell:untrusted');
+      }
+    }
     const dir = mutationProbeDir(context);
     fs.mkdirSync(dir, { recursive: true });
     const file = mutationProbeFile(context, key);
@@ -212,11 +228,12 @@ function registerMutationProbe(context, key, fingerprint, reserveIteration, obse
     const probe = {
       schemaVersion: 1,
       fingerprint,
-      reserveIteration: Boolean(reserveIteration),
-      countIteration: Boolean(reserveIteration),
+      reserveIteration: activeReservation,
+      countIteration: options.countIteration === undefined ? Boolean(reserveIteration) : Boolean(options.countIteration),
+      atLimitAllowance,
       observationCwd: observationCwd ? path.resolve(observationCwd) : null,
       observedAt,
-      reclaimAfterAt: observedAt + mutationProbeLeaseMs(timeoutMs),
+      reclaimAfterAt: observedAt + mutationProbeLeaseMs(options.timeoutMs),
     };
     atomicWriteJson(file, probe);
     return probe;

package/hooks/scripts/state-persist.js

@@ -3,14 +3,9 @@ const fs = require('fs');
 const path = require('path');
 const { getWorkspaceRoot, getSessionDir } = require('./lib/harness-state');
 const { loadWorkflow, saveWorkflow, isMajorWorkflow, peekMutationProbe, settleMutationProbe, settleMutationProbeConservative } = require('./lib/workflow-runtime');
-const { cwdOf, shellProbeKey, workspaceFingerprint, directMutationInWorkspace } = require('./lib/workflow-isolation');
+const { cwdOf, shellProbeKey, isVerificationShell, workspaceFingerprint, directMutationInWorkspace } = require('./lib/workflow-isolation');
 const { observeTool } = require('./lib/telemetry');
 
-// Commands that count as "verification ran" for the Stop gate
-// (hooks/scripts/stop-gate.js). Recall over precision: under-blocking the
-// gate is the correct failure direction, so a broad net is fine.
-const VERIFY_COMMAND_RE = /\b(test|spec|jest|vitest|mocha|pytest|rspec|phpunit|tsc|eslint|lint|build|compile|verify|check)\b/i;
-
 let inputData = '';
 const timeout = setTimeout(() => {
   processState(null);
@@ -127,7 +122,7 @@ function processState(payload) {
       }
       if ((toolName === 'Bash' || toolName === 'PowerShell') && !isFailed) {
         const command = (payload.tool_input && payload.tool_input.command) || '';
-        if (VERIFY_COMMAND_RE.test(command)) {
+        if (isVerificationShell(command)) {
           state.lastVerifyAt = Date.now();
           state.lastVerifyExitCode = exitCode ?? null;
         }

package/hooks/scripts/workflow-gate.js

@@ -4,7 +4,7 @@
 const path = require('path');
 const fs = require('fs');
 const { loadWorkflow, saveWorkflow, isMajorWorkflow, matchingRun, registerMutationProbe, recordBudgetEvent, readHookInput, WORKFLOW_CONTROLLER_COMMANDS } = require('./lib/workflow-runtime');
-const { key, classifyShell, workspaceFingerprint, shellProbeKey, cwdOf, commandOf, mutationPaths, directMutationInWorkspace, linkedWorktree, assertTargets, assertShellScope } = require('./lib/workflow-isolation');
+const { key, classifyShell, isVerificationShell, workspaceFingerprint, shellProbeKey, cwdOf, commandOf, mutationPaths, directMutationInWorkspace, linkedWorktree, assertTargets, assertShellScope } = require('./lib/workflow-isolation');
 
 const DIRECT = new Set(['Edit', 'Write', 'apply_patch']);
 const SHELL = new Set(['Bash', 'PowerShell', 'exec_command']);
@@ -88,14 +88,12 @@ function decide(payload) {
     return;
   }
   const probeKey = shellProbeKey(payload, cwd);
-  registerMutationProbe(
-    context,
-    probeKey,
-    fingerprint,
-    workflow.strategy === 'iterative-single',
-    cwd,
-    shellTimeoutMs(payload)
-  );
+  const iterative = workflow.strategy === 'iterative-single';
+  registerMutationProbe(context, probeKey, fingerprint, iterative, cwd, {
+    countIteration: iterative,
+    allowAtLimit: iterative && isVerificationShell(command),
+    timeoutMs: shellTimeoutMs(payload),
+  });
 }
 
 readHookInput(decide);

package/opencode-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "harness-enforcement",
   "description": "Repository inventory for the opencode enforcement plugin and its standalone helper",
-  "version": "0.12.1",
+  "version": "0.12.2",
   "runtime": "index.mjs",
   "hooks": {
     "verify": "hooks/verify.js"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "harness-everything",
-  "version": "0.12.1",
+  "version": "0.12.2",
   "description": "Harness OS: An Orchestrated Agent Operating System for AI-driven Development",
   "main": "harness-everything/scripts/bootstrap.js",
   "bin": {

package/plugins/harness-everything/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "harness-everything",
-  "version": "0.12.1",
+  "version": "0.12.2",
   "description": "Invariant-first engineering harness for routing, verification, and failure recovery while preserving agent autonomy.",
   "author": {
     "name": "Miya Daniel",

package/plugins/harness-everything/hooks/scripts/lib/workflow-isolation.js

@@ -135,6 +135,11 @@ function linkedWorktree(cwd, boundRoot) {
 // Unsupported shell syntax still fails closed.
 const DANGEROUS_FLAG_RE = /\s(?:--(?:output|ext-diff|textconv|exec|pre|pre-glob|pager|open|batch|filters)|-[xoO])(?:\b|=)|\s-(?:exec|execdir|delete|fprint|fprintf)\b/i;
 const READ_ONLY_PREFIX_RE = /^(?:git\s+(?:status|rev-parse|diff|log|show|ls-files|check-ignore)(?:\s|$)|git\s+branch\s+--show-current$|git\s+worktree\s+list(?:\s|$)|gh\s+(?:issue|pr)\s+(?:view|list)(?:\s|$)|gh\s+repo\s+view(?:\s|$)|gh\s+auth\s+status(?:\s|$)|(?:pwd|ls|dir|cat|type|head|tail|wc|stat|rg|grep|echo|Get-Location|Get-ChildItem|Get-Content|Select-String|Test-Path)(?:\s|$))/i;
+const VERIFY_COMMAND_RE = /\b(test|spec|jest|vitest|mocha|pytest|rspec|phpunit|tsc|eslint|lint|build|compile|verify|check)\b/i;
+
+function isVerificationShell(command) {
+  return VERIFY_COMMAND_RE.test(String(command || ''));
+}
 
 function isReadOnlySegment(segment) {
   return READ_ONLY_PREFIX_RE.test(segment) && !DANGEROUS_FLAG_RE.test(segment);
@@ -311,4 +316,4 @@ function assertShellScope(command, cwd, isolatedRoot) {
   }
 }
 
-module.exports = { canonical, key, within, linkedWorktree, classifyShell, workspaceFingerprint, shellProbeKey, inputOf, cwdOf, commandOf, mutationPaths, worktreeRoots, directMutationInWorkspace, assertTargets, assertShellScope };
+module.exports = { canonical, key, within, linkedWorktree, classifyShell, isVerificationShell, workspaceFingerprint, shellProbeKey, inputOf, cwdOf, commandOf, mutationPaths, worktreeRoots, directMutationInWorkspace, assertTargets, assertShellScope };

package/plugins/harness-everything/hooks/scripts/lib/workflow-runtime.js

@@ -196,14 +196,30 @@ function reclaimStaleMutationProbes(context, keepKey) {
   }
 }
 
-function registerMutationProbe(context, key, fingerprint, reserveIteration, observationCwd, timeoutMs) {
+function registerMutationProbe(context, key, fingerprint, reserveIteration, observationCwd, options = {}) {
   if (!MUTATION_PROBE_KEY.test(String(key || '')) || !MUTATION_PROBE_KEY.test(String(fingerprint || ''))) {
     throw new Error('invalid mutation probe identity');
   }
   return withMutationProbeLock(context, () => {
     refreshWorkflowForProbe(context);
     reclaimStaleMutationProbes(context, key);
-    if (reserveIteration) assertIterationCapacity(context, 'shell:untrusted');
+    let activeReservation = Boolean(reserveIteration);
+    let atLimitAllowance = false;
+    if (activeReservation) {
+      const budget = ensureWorkflowBudget(context.workflow);
+      const limit = budget.limits.maxIterations;
+      const reserved = mutationProbeReservations(context);
+      if (options.allowAtLimit === true && budget.state === 'active' && limit !== null &&
+          budget.counters.iterations === limit && reserved === 0) {
+        // Verification-shaped commands get observation-only admission at the
+        // exact mutation limit. If they mutate, settlement still records the
+        // mutation and exhausts the budget after execution (#164).
+        activeReservation = false;
+        atLimitAllowance = true;
+      } else {
+        assertIterationCapacity(context, 'shell:untrusted');
+      }
+    }
     const dir = mutationProbeDir(context);
     fs.mkdirSync(dir, { recursive: true });
     const file = mutationProbeFile(context, key);
@@ -212,11 +228,12 @@ function registerMutationProbe(context, key, fingerprint, reserveIteration, obse
     const probe = {
       schemaVersion: 1,
       fingerprint,
-      reserveIteration: Boolean(reserveIteration),
-      countIteration: Boolean(reserveIteration),
+      reserveIteration: activeReservation,
+      countIteration: options.countIteration === undefined ? Boolean(reserveIteration) : Boolean(options.countIteration),
+      atLimitAllowance,
       observationCwd: observationCwd ? path.resolve(observationCwd) : null,
       observedAt,
-      reclaimAfterAt: observedAt + mutationProbeLeaseMs(timeoutMs),
+      reclaimAfterAt: observedAt + mutationProbeLeaseMs(options.timeoutMs),
     };
     atomicWriteJson(file, probe);
     return probe;

package/plugins/harness-everything/hooks/scripts/state-persist.js

@@ -3,14 +3,9 @@ const fs = require('fs');
 const path = require('path');
 const { getWorkspaceRoot, getSessionDir } = require('./lib/harness-state');
 const { loadWorkflow, saveWorkflow, isMajorWorkflow, peekMutationProbe, settleMutationProbe, settleMutationProbeConservative } = require('./lib/workflow-runtime');
-const { cwdOf, shellProbeKey, workspaceFingerprint, directMutationInWorkspace } = require('./lib/workflow-isolation');
+const { cwdOf, shellProbeKey, isVerificationShell, workspaceFingerprint, directMutationInWorkspace } = require('./lib/workflow-isolation');
 const { observeTool } = require('./lib/telemetry');
 
-// Commands that count as "verification ran" for the Stop gate
-// (hooks/scripts/stop-gate.js). Recall over precision: under-blocking the
-// gate is the correct failure direction, so a broad net is fine.
-const VERIFY_COMMAND_RE = /\b(test|spec|jest|vitest|mocha|pytest|rspec|phpunit|tsc|eslint|lint|build|compile|verify|check)\b/i;
-
 let inputData = '';
 const timeout = setTimeout(() => {
   processState(null);
@@ -127,7 +122,7 @@ function processState(payload) {
       }
       if ((toolName === 'Bash' || toolName === 'PowerShell') && !isFailed) {
         const command = (payload.tool_input && payload.tool_input.command) || '';
-        if (VERIFY_COMMAND_RE.test(command)) {
+        if (isVerificationShell(command)) {
           state.lastVerifyAt = Date.now();
           state.lastVerifyExitCode = exitCode ?? null;
         }

package/plugins/harness-everything/hooks/scripts/workflow-gate.js

@@ -4,7 +4,7 @@
 const path = require('path');
 const fs = require('fs');
 const { loadWorkflow, saveWorkflow, isMajorWorkflow, matchingRun, registerMutationProbe, recordBudgetEvent, readHookInput, WORKFLOW_CONTROLLER_COMMANDS } = require('./lib/workflow-runtime');
-const { key, classifyShell, workspaceFingerprint, shellProbeKey, cwdOf, commandOf, mutationPaths, directMutationInWorkspace, linkedWorktree, assertTargets, assertShellScope } = require('./lib/workflow-isolation');
+const { key, classifyShell, isVerificationShell, workspaceFingerprint, shellProbeKey, cwdOf, commandOf, mutationPaths, directMutationInWorkspace, linkedWorktree, assertTargets, assertShellScope } = require('./lib/workflow-isolation');
 
 const DIRECT = new Set(['Edit', 'Write', 'apply_patch']);
 const SHELL = new Set(['Bash', 'PowerShell', 'exec_command']);
@@ -88,14 +88,12 @@ function decide(payload) {
     return;
   }
   const probeKey = shellProbeKey(payload, cwd);
-  registerMutationProbe(
-    context,
-    probeKey,
-    fingerprint,
-    workflow.strategy === 'iterative-single',
-    cwd,
-    shellTimeoutMs(payload)
-  );
+  const iterative = workflow.strategy === 'iterative-single';
+  registerMutationProbe(context, probeKey, fingerprint, iterative, cwd, {
+    countIteration: iterative,
+    allowAtLimit: iterative && isVerificationShell(command),
+    timeoutMs: shellTimeoutMs(payload),
+  });
 }
 
 readHookInput(decide);

package/plugins/harness-everything/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
   "name": "harness-everything",
-  "version": "0.12.1",
+  "version": "0.12.2",
   "description": "Invariant-first engineering harness for routing, verification, and failure recovery while preserving agent autonomy.",
   "author": {
     "name": "Miya Daniel",