anvil-engineering-workbench 1.0.9 → 1.1.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (162) hide show
  1. package/CHANGELOG.md +31 -1
  2. package/README.md +23 -18
  3. package/dist/ai/claude-code-provider.js +55 -9
  4. package/dist/ai/claude-code-provider.js.map +1 -1
  5. package/dist/ai/provider.js +1 -1
  6. package/dist/ai/provider.js.map +1 -1
  7. package/dist/cli.js +837 -278
  8. package/dist/cli.js.map +1 -1
  9. package/dist/core/adr-engine.js +9 -11
  10. package/dist/core/adr-engine.js.map +1 -1
  11. package/dist/core/agent-config-release.js +29 -13
  12. package/dist/core/agent-config-release.js.map +1 -1
  13. package/dist/core/agent-config-versioning.js +32 -17
  14. package/dist/core/agent-config-versioning.js.map +1 -1
  15. package/dist/core/agent-evaluation-intelligence.js +30 -38
  16. package/dist/core/agent-evaluation-intelligence.js.map +1 -1
  17. package/dist/core/agent-evolution.js +109 -20
  18. package/dist/core/agent-evolution.js.map +1 -1
  19. package/dist/core/agent-experiment-studio.js +19 -11
  20. package/dist/core/agent-experiment-studio.js.map +1 -1
  21. package/dist/core/agent-optimization.js +69 -14
  22. package/dist/core/agent-optimization.js.map +1 -1
  23. package/dist/core/agent-registry.js +87 -15
  24. package/dist/core/agent-registry.js.map +1 -1
  25. package/dist/core/agent-security-audit.js +21 -21
  26. package/dist/core/agent-security-audit.js.map +1 -1
  27. package/dist/core/agent-workspace-context.js +37 -6
  28. package/dist/core/agent-workspace-context.js.map +1 -1
  29. package/dist/core/ai-code-review.js +60 -12
  30. package/dist/core/ai-code-review.js.map +1 -1
  31. package/dist/core/anthropic-provider.js.map +1 -1
  32. package/dist/core/architecture-map.js +87 -24
  33. package/dist/core/architecture-map.js.map +1 -1
  34. package/dist/core/benchmark-engine.js +89 -21
  35. package/dist/core/benchmark-engine.js.map +1 -1
  36. package/dist/core/benchmark-matrix.js +215 -41
  37. package/dist/core/benchmark-matrix.js.map +1 -1
  38. package/dist/core/benchmark-regression.js +124 -18
  39. package/dist/core/benchmark-regression.js.map +1 -1
  40. package/dist/core/change-intelligence.js +102 -43
  41. package/dist/core/change-intelligence.js.map +1 -1
  42. package/dist/core/code-intelligence.js +40 -13
  43. package/dist/core/code-intelligence.js.map +1 -1
  44. package/dist/core/code-review.js +145 -17
  45. package/dist/core/code-review.js.map +1 -1
  46. package/dist/core/code-search.js +79 -39
  47. package/dist/core/code-search.js.map +1 -1
  48. package/dist/core/container-sandbox.js +87 -23
  49. package/dist/core/container-sandbox.js.map +1 -1
  50. package/dist/core/dashboard-server.js +379 -93
  51. package/dist/core/dashboard-server.js.map +1 -1
  52. package/dist/core/dependency-graph.js +75 -13
  53. package/dist/core/dependency-graph.js.map +1 -1
  54. package/dist/core/dependency-impact.js +132 -33
  55. package/dist/core/dependency-impact.js.map +1 -1
  56. package/dist/core/dependency-intelligence.js +153 -56
  57. package/dist/core/dependency-intelligence.js.map +1 -1
  58. package/dist/core/doctor.js +3 -3
  59. package/dist/core/doctor.js.map +1 -1
  60. package/dist/core/engineering-risk.js +75 -40
  61. package/dist/core/engineering-risk.js.map +1 -1
  62. package/dist/core/env-compat.js.map +1 -1
  63. package/dist/core/evaluation-engine.js +188 -42
  64. package/dist/core/evaluation-engine.js.map +1 -1
  65. package/dist/core/execution-trace.js +83 -29
  66. package/dist/core/execution-trace.js.map +1 -1
  67. package/dist/core/experiment-history.js +70 -20
  68. package/dist/core/experiment-history.js.map +1 -1
  69. package/dist/core/experiment-manifest.js +92 -13
  70. package/dist/core/experiment-manifest.js.map +1 -1
  71. package/dist/core/gemini-provider.js.map +1 -1
  72. package/dist/core/git-intelligence.js +68 -16
  73. package/dist/core/git-intelligence.js.map +1 -1
  74. package/dist/core/hook-engine.js +7 -3
  75. package/dist/core/hook-engine.js.map +1 -1
  76. package/dist/core/hosted-agent-provider.js +48 -8
  77. package/dist/core/hosted-agent-provider.js.map +1 -1
  78. package/dist/core/intelligence-timeline.js +92 -43
  79. package/dist/core/intelligence-timeline.js.map +1 -1
  80. package/dist/core/mcp-setup.js +12 -12
  81. package/dist/core/mcp-setup.js.map +1 -1
  82. package/dist/core/model-router.js +153 -37
  83. package/dist/core/model-router.js.map +1 -1
  84. package/dist/core/openai-compatible-provider.js +41 -10
  85. package/dist/core/openai-compatible-provider.js.map +1 -1
  86. package/dist/core/orchestrator.js +28 -23
  87. package/dist/core/orchestrator.js.map +1 -1
  88. package/dist/core/patch-application.js +88 -24
  89. package/dist/core/patch-application.js.map +1 -1
  90. package/dist/core/patch-approval.js +43 -15
  91. package/dist/core/patch-approval.js.map +1 -1
  92. package/dist/core/patch-validation.js +93 -15
  93. package/dist/core/patch-validation.js.map +1 -1
  94. package/dist/core/permission-policy.js +8 -5
  95. package/dist/core/permission-policy.js.map +1 -1
  96. package/dist/core/policy-engine.d.ts +76 -0
  97. package/dist/core/policy-engine.js +352 -0
  98. package/dist/core/policy-engine.js.map +1 -0
  99. package/dist/core/project-cleaner.js +143 -129
  100. package/dist/core/project-cleaner.js.map +1 -1
  101. package/dist/core/project-health.js +183 -21
  102. package/dist/core/project-health.js.map +1 -1
  103. package/dist/core/project-init.js +32 -20
  104. package/dist/core/project-init.js.map +1 -1
  105. package/dist/core/project-inspector.js +81 -32
  106. package/dist/core/project-inspector.js.map +1 -1
  107. package/dist/core/project-memory.js +41 -18
  108. package/dist/core/project-memory.js.map +1 -1
  109. package/dist/core/provider-diagnostics.js +65 -12
  110. package/dist/core/provider-diagnostics.js.map +1 -1
  111. package/dist/core/quality-audit.js +167 -28
  112. package/dist/core/quality-audit.js.map +1 -1
  113. package/dist/core/regression-intelligence.js +88 -8
  114. package/dist/core/regression-intelligence.js.map +1 -1
  115. package/dist/core/remediation-planner.js +112 -40
  116. package/dist/core/remediation-planner.js.map +1 -1
  117. package/dist/core/review-history.js +39 -7
  118. package/dist/core/review-history.js.map +1 -1
  119. package/dist/core/run-history-output.js.map +1 -1
  120. package/dist/core/run-replay.js +12 -10
  121. package/dist/core/run-replay.js.map +1 -1
  122. package/dist/core/run-store.js +48 -21
  123. package/dist/core/run-store.js.map +1 -1
  124. package/dist/core/sandbox-cleanup-audit.js +1 -1
  125. package/dist/core/sandbox-cleanup-audit.js.map +1 -1
  126. package/dist/core/sandbox-compliance.js +53 -9
  127. package/dist/core/sandbox-compliance.js.map +1 -1
  128. package/dist/core/sandbox-lifecycle.js +1 -1
  129. package/dist/core/sandbox-lifecycle.js.map +1 -1
  130. package/dist/core/sandbox-policy.js +27 -3
  131. package/dist/core/sandbox-policy.js.map +1 -1
  132. package/dist/core/sandbox-runtime.js +9 -5
  133. package/dist/core/sandbox-runtime.js.map +1 -1
  134. package/dist/core/sandbox-security-profiles.js +31 -12
  135. package/dist/core/sandbox-security-profiles.js.map +1 -1
  136. package/dist/core/sarif-formatter.d.ts +10 -3
  137. package/dist/core/sarif-formatter.js +172 -57
  138. package/dist/core/sarif-formatter.js.map +1 -1
  139. package/dist/core/security-scanner.js +40 -32
  140. package/dist/core/security-scanner.js.map +1 -1
  141. package/dist/core/skill-registry.js +83 -21
  142. package/dist/core/skill-registry.js.map +1 -1
  143. package/dist/core/staged-verification.js +25 -10
  144. package/dist/core/staged-verification.js.map +1 -1
  145. package/dist/core/verification-history.js +12 -7
  146. package/dist/core/verification-history.js.map +1 -1
  147. package/dist/core/verification.js +50 -19
  148. package/dist/core/verification.js.map +1 -1
  149. package/dist/core/workflow-store.js +16 -4
  150. package/dist/core/workflow-store.js.map +1 -1
  151. package/dist/core/workspace-browser.js +57 -10
  152. package/dist/core/workspace-browser.js.map +1 -1
  153. package/dist/core/workspace-scan-policy.js +26 -4
  154. package/dist/core/workspace-scan-policy.js.map +1 -1
  155. package/dist/core/workspace-tool-broker.d.ts +1 -1
  156. package/dist/core/workspace-tool-broker.js +25 -4
  157. package/dist/core/workspace-tool-broker.js.map +1 -1
  158. package/dist/mcp/server.js +148 -55
  159. package/dist/mcp/server.js.map +1 -1
  160. package/dist/mcp/workspace-path.js +2 -2
  161. package/dist/mcp/workspace-path.js.map +1 -1
  162. package/package.json +8 -5
package/dist/cli.js CHANGED
@@ -10,7 +10,7 @@ import { formatArchitectureMap, mapArchitecture } from "./core/architecture-map.
10
10
  import { formatProjectReport, inspectProject } from "./core/project-inspector.js";
11
11
  import { loadAgentRegistry } from "./core/agent-registry.js";
12
12
  import { createReviewPlan } from "./core/review-planner.js";
13
- import { saveRun, listRuns, loadRun, decideTaskApproval, recoverInterruptedRun, withRunExecutionLock } from "./core/run-store.js";
13
+ import { saveRun, listRuns, loadRun, decideTaskApproval, recoverInterruptedRun, withRunExecutionLock, } from "./core/run-store.js";
14
14
  import { advanceRun } from "./core/orchestrator.js";
15
15
  import { formatRunHistoryJson } from "./core/run-history-output.js";
16
16
  import { OpenAICompatibleProvider } from "./core/openai-compatible-provider.js";
@@ -19,7 +19,7 @@ import { ClaudeCodeProvider } from "./ai/claude-code-provider.js";
19
19
  import { startDashboard } from "./core/dashboard-server.js";
20
20
  import { auditProject, formatQualityAudit } from "./core/quality-audit.js";
21
21
  import { discoverChecks, formatVerificationResult, runVerification } from "./core/verification.js";
22
- import { compareVerification, formatVerificationComparison, listVerifications, saveVerification } from "./core/verification-history.js";
22
+ import { compareVerification, formatVerificationComparison, listVerifications, saveVerification, } from "./core/verification-history.js";
23
23
  import { formatProjectHealthMarkdown, generateProjectHealth } from "./core/project-health.js";
24
24
  import { analyzeCodebase, formatCodebaseAnalysis } from "./core/code-intelligence.js";
25
25
  import { analyzeGitChanges, formatGitIntelligence } from "./core/git-intelligence.js";
@@ -32,45 +32,46 @@ import { stageApprovedPatch, reviewStagedPatch } from "./core/patch-application.
32
32
  import { verifyStagedPatch } from "./core/staged-verification.js";
33
33
  import { formatSandboxRuntimeReport, parseSandboxRuntimeReport } from "./core/sandbox-runtime.js";
34
34
  import { detectDockerRootless } from "./core/container-sandbox.js";
35
- import { formatSandboxCleanupReport, parseManagedContainers, selectStoppedManagedContainers } from "./core/sandbox-lifecycle.js";
35
+ import { formatSandboxCleanupReport, parseManagedContainers, selectStoppedManagedContainers, } from "./core/sandbox-lifecycle.js";
36
36
  import { getSandboxSecurityProfile, listSandboxSecurityProfiles } from "./core/sandbox-security-profiles.js";
37
37
  import { createSandboxComplianceReport, formatSandboxComplianceReport } from "./core/sandbox-compliance.js";
38
38
  import { formatSandboxPolicyValidation, loadSandboxPolicy } from "./core/sandbox-policy.js";
39
- import { appendSandboxCleanupAudit, formatSandboxCleanupHistory, readSandboxCleanupAudit } from "./core/sandbox-cleanup-audit.js";
39
+ import { appendSandboxCleanupAudit, formatSandboxCleanupHistory, readSandboxCleanupAudit, } from "./core/sandbox-cleanup-audit.js";
40
40
  import { formatRegressionReport, generateRegressionReport } from "./core/regression-intelligence.js";
41
41
  import { searchWorkspace } from "./core/code-search.js";
42
42
  import { buildAgentWorkspaceContext } from "./core/agent-workspace-context.js";
43
43
  import { analyzeDependencyImpact, formatDependencyImpact } from "./core/dependency-impact.js";
44
44
  import { analyzeEngineeringRisk, formatEngineeringRisk } from "./core/engineering-risk.js";
45
45
  import { analyzeChangeIntelligence, formatChangeIntelligence } from "./core/change-intelligence.js";
46
- import { captureIntelligenceSnapshot, compareLatestIntelligence, formatIntelligenceComparison } from "./core/intelligence-timeline.js";
46
+ import { captureIntelligenceSnapshot, compareLatestIntelligence, formatIntelligenceComparison, } from "./core/intelligence-timeline.js";
47
47
  import { analyzeDependencies, formatDependencyIntelligence } from "./core/dependency-intelligence.js";
48
48
  import { generateRemediationPlan, formatRemediationPlan } from "./core/remediation-planner.js";
49
- import { addProjectMemory, buildProjectMemoryContext, listProjectMemory, removeProjectMemory } from "./core/project-memory.js";
49
+ import { addProjectMemory, buildProjectMemoryContext, listProjectMemory, removeProjectMemory, } from "./core/project-memory.js";
50
50
  import { setupHarnessMcp, resolveMcpServerScript } from "./core/mcp-setup.js";
51
51
  import { initAdrDirectory, listAdrs, createAdr, syncMemoryToAdrs } from "./core/adr-engine.js";
52
52
  import { scanProjectSecurity, applySecurityFixes, formatSecurityReport } from "./core/security-scanner.js";
53
+ import { evaluateGate, generateBaselineJson } from "./core/policy-engine.js";
53
54
  import { formatSkillContext, loadSkillRegistry, selectSkills } from "./core/skill-registry.js";
54
55
  import { applyBeforeAgentHooks, loadHooks } from "./core/hook-engine.js";
55
56
  import { assertSupportedAgentPolicy, loadPermissionPolicy } from "./core/permission-policy.js";
56
57
  import { ModelRouter } from "./core/model-router.js";
57
58
  import { loadExecutionTrace, saveExecutionTrace } from "./core/execution-trace.js";
58
59
  import { createReplayRun } from "./core/run-replay.js";
59
- import { evaluateRun, evaluateRunWithJudge, formatEvaluationReport, loadEvaluationSuites, saveEvaluationReport, loadEvaluationReport } from "./core/evaluation-engine.js";
60
- import { createBenchmarkTrial, formatBenchmark, listBenchmarks, recordBenchmarkTrial } from "./core/benchmark-engine.js";
61
- import { executeBenchmarkMatrix, formatBenchmarkMatrix, loadBenchmarkMatrix, listBenchmarkMatrices, saveBenchmarkMatrix } from "./core/benchmark-matrix.js";
60
+ import { evaluateRun, evaluateRunWithJudge, formatEvaluationReport, loadEvaluationSuites, saveEvaluationReport, loadEvaluationReport, } from "./core/evaluation-engine.js";
61
+ import { createBenchmarkTrial, formatBenchmark, listBenchmarks, recordBenchmarkTrial, } from "./core/benchmark-engine.js";
62
+ import { executeBenchmarkMatrix, formatBenchmarkMatrix, loadBenchmarkMatrix, listBenchmarkMatrices, saveBenchmarkMatrix, } from "./core/benchmark-matrix.js";
62
63
  import { fingerprintAgentConfig } from "./core/benchmark-matrix.js";
63
64
  import { loadAgentConfigVersion } from "./core/agent-config-versioning.js";
64
65
  import { formatBenchmarkRegression, loadBenchmarkRegression } from "./core/benchmark-regression.js";
65
66
  import { buildAgentOptimizationPlan, formatAgentOptimizationPlan } from "./core/agent-optimization.js";
66
67
  import { auditAgentConfiguration, formatAgentSecurityReport } from "./core/agent-security-audit.js";
67
- import { buildAgentEvaluationIntelligence, formatAgentEvaluationIntelligence } from "./core/agent-evaluation-intelligence.js";
68
+ import { buildAgentEvaluationIntelligence, formatAgentEvaluationIntelligence, } from "./core/agent-evaluation-intelligence.js";
68
69
  import { assessAgentEvolution, formatAgentEvolutionAssessment } from "./core/agent-evolution.js";
69
- import { createSkillExperimentMatrix, formatSkillExperimentMatrix, createModelComparisonMatrix, formatModelComparisonMatrix } from "./core/agent-experiment-studio.js";
70
+ import { createSkillExperimentMatrix, formatSkillExperimentMatrix, createModelComparisonMatrix, formatModelComparisonMatrix, } from "./core/agent-experiment-studio.js";
70
71
  import { buildExperimentHistory, formatExperimentHistory } from "./core/experiment-history.js";
71
- import { captureAgentConfigVersion, compareAgentConfigVersions, formatAgentConfigVersion, listAgentConfigVersions, materializeAgentConfigVersion } from "./core/agent-config-versioning.js";
72
- import { decideAgentConfigRelease, formatAgentConfigRelease, listAgentConfigReleases, prepareAgentConfigRelease } from "./core/agent-config-release.js";
73
- import { captureExperimentManifest, compareExperimentManifests, formatExperimentManifest, loadExperimentManifest, saveExperimentManifest } from "./core/experiment-manifest.js";
72
+ import { captureAgentConfigVersion, compareAgentConfigVersions, formatAgentConfigVersion, listAgentConfigVersions, materializeAgentConfigVersion, } from "./core/agent-config-versioning.js";
73
+ import { decideAgentConfigRelease, formatAgentConfigRelease, listAgentConfigReleases, prepareAgentConfigRelease, } from "./core/agent-config-release.js";
74
+ import { captureExperimentManifest, compareExperimentManifests, formatExperimentManifest, loadExperimentManifest, saveExperimentManifest, } from "./core/experiment-manifest.js";
74
75
  function runSandboxProbe(args) {
75
76
  return new Promise((resolve, reject) => {
76
77
  const child = spawn("docker", args, { shell: false, windowsHide: true, stdio: ["ignore", "pipe", "pipe"] });
@@ -92,8 +93,8 @@ function runSandboxProbe(args) {
92
93
  };
93
94
  child.stdout.on("data", append);
94
95
  child.stderr.on("data", append);
95
- child.once("error", error => finish(error));
96
- child.once("close", code => code === 0 ? finish() : finish(new Error("Docker probe exited with status " + code + ".")));
96
+ child.once("error", (error) => finish(error));
97
+ child.once("close", (code) => code === 0 ? finish() : finish(new Error("Docker probe exited with status " + code + ".")));
97
98
  const timer = setTimeout(() => {
98
99
  child.kill("SIGTERM");
99
100
  finish(new Error("Docker probe timed out after 15 seconds."));
@@ -247,7 +248,7 @@ async function main(args) {
247
248
  process.stdout.write(`anvil ${VERSION}\n`);
248
249
  return;
249
250
  case "init": {
250
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
251
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
251
252
  try {
252
253
  const result = await initializeProject(path.resolve(target));
253
254
  process.stdout.write(`ANVIL initialized in ${path.resolve(target)}\n`);
@@ -264,10 +265,13 @@ async function main(args) {
264
265
  return;
265
266
  }
266
267
  case "evals": {
267
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
268
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
268
269
  try {
269
270
  const suites = await loadEvaluationSuites(target);
270
- process.stdout.write(suites.length ? suites.map(suite => suite.id + " " + suite.name + " · " + suite.cases.length + " cases").join("\n") + "\n" : "No evaluation suites. Add JSON suites under .forge/evals/.\n");
271
+ process.stdout.write(suites.length
272
+ ? suites.map((suite) => suite.id + " " + suite.name + " · " + suite.cases.length + " cases").join("\n") +
273
+ "\n"
274
+ : "No evaluation suites. Add JSON suites under .forge/evals/.\n");
271
275
  }
272
276
  catch (error) {
273
277
  process.stderr.write("Could not load evaluation suites: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -277,10 +281,18 @@ async function main(args) {
277
281
  }
278
282
  case "agent-release": {
279
283
  const action = rest[0] ?? "list";
280
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
284
+ const option = (name) => {
285
+ const index = rest.indexOf(name);
286
+ return index >= 0 ? rest[index + 1] : undefined;
287
+ };
281
288
  const pathIndex = rest.indexOf("--path");
282
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
283
- const args = rest.slice(1).filter((arg, index, all) => !arg.startsWith("--") && all[index - 1] !== "--path" && all[index - 1] !== "--reviewer" && all[index - 1] !== "--matrix");
289
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
290
+ const args = rest
291
+ .slice(1)
292
+ .filter((arg, index, all) => !arg.startsWith("--") &&
293
+ all[index - 1] !== "--path" &&
294
+ all[index - 1] !== "--reviewer" &&
295
+ all[index - 1] !== "--matrix");
284
296
  try {
285
297
  if (action === "prepare") {
286
298
  const [candidateVersionId, baselineVersionId] = args;
@@ -288,18 +300,32 @@ async function main(args) {
288
300
  if (!candidateVersionId || !baselineVersionId || !matrixId)
289
301
  throw new Error("Usage: anvil agent-release prepare CANDIDATE_CONFIG_ID BASELINE_CONFIG_ID --matrix MATRIX_ID --path PROJECT");
290
302
  const regression = await loadBenchmarkRegression(target, matrixId);
291
- const release = await prepareAgentConfigRelease(target, { candidateVersionId, baselineVersionId, regression });
303
+ const release = await prepareAgentConfigRelease(target, {
304
+ candidateVersionId,
305
+ baselineVersionId,
306
+ regression,
307
+ });
292
308
  process.stdout.write(formatAgentConfigRelease(release) + "\nHuman approval is still required. No live files were changed.\n");
293
309
  }
294
310
  else if (action === "list") {
295
311
  const items = await listAgentConfigReleases(target);
296
- process.stdout.write(items.length ? items.map(item => item.id + " " + item.status.padEnd(17) + " " + item.candidateVersionId + " " + item.benchmarkStatus).join("\n") + "\n" : "No agent configuration releases found.\n");
312
+ process.stdout.write(items.length
313
+ ? items
314
+ .map((item) => item.id +
315
+ " " +
316
+ item.status.padEnd(17) +
317
+ " " +
318
+ item.candidateVersionId +
319
+ " " +
320
+ item.benchmarkStatus)
321
+ .join("\n") + "\n"
322
+ : "No agent configuration releases found.\n");
297
323
  }
298
324
  else if (action === "show") {
299
325
  const id = args[0];
300
326
  if (!id)
301
327
  throw new Error("Usage: anvil agent-release show RELEASE_ID --path PROJECT");
302
- const item = (await listAgentConfigReleases(target)).find(entry => entry.id === id);
328
+ const item = (await listAgentConfigReleases(target)).find((entry) => entry.id === id);
303
329
  if (!item)
304
330
  throw new Error("Release not found: " + id);
305
331
  process.stdout.write(formatAgentConfigRelease(item) + "\n");
@@ -326,17 +352,39 @@ async function main(args) {
326
352
  case "agent-config": {
327
353
  const action = rest[0] ?? "list";
328
354
  const pathIndex = rest.indexOf("--path");
329
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
330
- const ids = rest.slice(1).filter((arg, index, all) => !arg.startsWith("--") && all[index - 1] !== "--path" && all[index - 1] !== "--note" && all[index - 1] !== "--reviewer");
331
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
355
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
356
+ const ids = rest
357
+ .slice(1)
358
+ .filter((arg, index, all) => !arg.startsWith("--") &&
359
+ all[index - 1] !== "--path" &&
360
+ all[index - 1] !== "--note" &&
361
+ all[index - 1] !== "--reviewer");
362
+ const option = (name) => {
363
+ const index = rest.indexOf(name);
364
+ return index >= 0 ? rest[index + 1] : undefined;
365
+ };
332
366
  try {
333
367
  if (action === "snapshot") {
334
- const version = await captureAgentConfigVersion(target, { note: option("--note") ?? "", createdBy: option("--author") ?? "local-user" });
368
+ const version = await captureAgentConfigVersion(target, {
369
+ note: option("--note") ?? "",
370
+ createdBy: option("--author") ?? "local-user",
371
+ });
335
372
  process.stdout.write(formatAgentConfigVersion(version) + "\n");
336
373
  }
337
374
  else if (action === "list") {
338
375
  const versions = await listAgentConfigVersions(target);
339
- process.stdout.write(versions.length ? versions.map(item => item.id + " " + item.status.padEnd(9) + " " + item.createdAt + " " + item.files.length + " files").join("\n") + "\n" : "No agent configuration snapshots found. Run forge agent-config snapshot first.\n");
376
+ process.stdout.write(versions.length
377
+ ? versions
378
+ .map((item) => item.id +
379
+ " " +
380
+ item.status.padEnd(9) +
381
+ " " +
382
+ item.createdAt +
383
+ " " +
384
+ item.files.length +
385
+ " files")
386
+ .join("\n") + "\n"
387
+ : "No agent configuration snapshots found. Run forge agent-config snapshot first.\n");
340
388
  }
341
389
  else if (action === "show") {
342
390
  const id = ids[0];
@@ -347,9 +395,18 @@ async function main(args) {
347
395
  else if (action === "diff") {
348
396
  if (ids.length < 2)
349
397
  throw new Error("Usage: anvil agent-config diff BEFORE_ID AFTER_ID --path PROJECT");
350
- const [before, after] = await Promise.all([loadAgentConfigVersion(target, ids[0]), loadAgentConfigVersion(target, ids[1])]);
398
+ const [before, after] = await Promise.all([
399
+ loadAgentConfigVersion(target, ids[0]),
400
+ loadAgentConfigVersion(target, ids[1]),
401
+ ]);
351
402
  const diff = compareAgentConfigVersions(before, after);
352
- process.stdout.write(["Agent configuration diff", "Added: " + (diff.added.join(", ") || "none"), "Changed: " + (diff.changed.join(", ") || "none"), "Removed: " + (diff.removed.join(", ") || "none"), "Unchanged: " + diff.unchanged.length].join("\n") + "\n");
403
+ process.stdout.write([
404
+ "Agent configuration diff",
405
+ "Added: " + (diff.added.join(", ") || "none"),
406
+ "Changed: " + (diff.changed.join(", ") || "none"),
407
+ "Removed: " + (diff.removed.join(", ") || "none"),
408
+ "Unchanged: " + diff.unchanged.length,
409
+ ].join("\n") + "\n");
353
410
  }
354
411
  else if (action === "approve") {
355
412
  throw new Error("Direct snapshot approval is disabled. Use forge agent-release prepare with a release-ready benchmark matrix, then forge agent-release approve RELEASE_ID --reviewer NAME.");
@@ -364,9 +421,12 @@ async function main(args) {
364
421
  return;
365
422
  }
366
423
  case "experiment-history": {
367
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
424
+ const option = (name) => {
425
+ const index = rest.indexOf(name);
426
+ return index >= 0 ? rest[index + 1] : undefined;
427
+ };
368
428
  const pathIndex = rest.indexOf("--path");
369
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
429
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
370
430
  const rawLimit = option("--limit");
371
431
  const limit = rawLimit === undefined ? undefined : Number(rawLimit);
372
432
  if (limit !== undefined && (!Number.isSafeInteger(limit) || limit < 1 || limit > 100)) {
@@ -378,7 +438,11 @@ async function main(args) {
378
438
  const suiteId = option("--suite");
379
439
  const candidateId = option("--candidate");
380
440
  const reports = await listBenchmarkMatrices(target);
381
- const history = buildExperimentHistory(reports, { ...(suiteId ? { suiteId } : {}), ...(candidateId ? { candidateId } : {}), ...(limit !== undefined ? { limit } : {}) });
441
+ const history = buildExperimentHistory(reports, {
442
+ ...(suiteId ? { suiteId } : {}),
443
+ ...(candidateId ? { candidateId } : {}),
444
+ ...(limit !== undefined ? { limit } : {}),
445
+ });
382
446
  process.stdout.write(rest.includes("--json") ? JSON.stringify(history, null, 2) + "\n" : formatExperimentHistory(history) + "\n");
383
447
  }
384
448
  catch (error) {
@@ -389,7 +453,10 @@ async function main(args) {
389
453
  }
390
454
  case "experiment-studio": {
391
455
  const action = rest[0] ?? "create";
392
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
456
+ const option = (name) => {
457
+ const index = rest.indexOf(name);
458
+ return index >= 0 ? rest[index + 1] : undefined;
459
+ };
393
460
  if (action !== "create" && action !== "compare") {
394
461
  process.stderr.write("Usage: anvil experiment-studio compare --id ID --name NAME --candidates id:provider:model,id:provider:model [--config-versions id=cfg-id,id=cfg-id] [--repeats 2] [--path PROJECT]\n");
395
462
  process.stderr.write(" or: forge experiment-studio create --id ID --name NAME --provider PROVIDER --model MODEL --skills skill-a,skill-b [--repeats 2] [--path PROJECT]\n");
@@ -402,20 +469,23 @@ async function main(args) {
402
469
  const rawCandidates = option("--candidates");
403
470
  const repeats = Number(option("--repeats") ?? "2");
404
471
  const pathIndex = rest.indexOf("--path");
405
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
472
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
406
473
  if (!id || !name || !rawCandidates) {
407
474
  process.stderr.write("Usage: anvil experiment-studio compare --id ID --name NAME --candidates id:provider:model,id:provider:model [--repeats 2] [--path PROJECT]\n");
408
475
  process.exitCode = 1;
409
476
  return;
410
477
  }
411
478
  try {
412
- const configVersions = new Map((option("--config-versions") ?? "").split(",").filter(Boolean).map(item => {
479
+ const configVersions = new Map((option("--config-versions") ?? "")
480
+ .split(",")
481
+ .filter(Boolean)
482
+ .map((item) => {
413
483
  const separator = item.indexOf("=");
414
484
  if (separator < 1 || separator === item.length - 1)
415
485
  throw new Error("Each config version must use candidate-id=cfg-xxxxxxxxxxxx format.");
416
486
  return [item.slice(0, separator), item.slice(separator + 1)];
417
487
  }));
418
- const candidates = rawCandidates.split(",").map(item => {
488
+ const candidates = rawCandidates.split(",").map((item) => {
419
489
  const [candidateId, candidateProvider, ...modelParts] = item.split(":");
420
490
  const model = modelParts.join(":");
421
491
  if (!candidateId || !candidateProvider || !model)
@@ -423,10 +493,15 @@ async function main(args) {
423
493
  if (!["openai", "claude-code", "anthropic", "google", "ollama"].includes(candidateProvider))
424
494
  throw new Error("Unsupported candidate provider: " + candidateProvider);
425
495
  const configVersionId = configVersions.get(candidateId);
426
- return { id: candidateId, provider: candidateProvider, model, ...(configVersionId ? { configVersionId } : {}) };
496
+ return {
497
+ id: candidateId,
498
+ provider: candidateProvider,
499
+ model,
500
+ ...(configVersionId ? { configVersionId } : {}),
501
+ };
427
502
  });
428
503
  for (const candidateId of configVersions.keys())
429
- if (!candidates.some(candidate => candidate.id === candidateId))
504
+ if (!candidates.some((candidate) => candidate.id === candidateId))
430
505
  throw new Error("Configuration snapshot references unknown candidate: " + candidateId);
431
506
  for (const candidate of candidates)
432
507
  if (candidate.configVersionId)
@@ -435,7 +510,11 @@ async function main(args) {
435
510
  const destination = path.join(target, ".forge", "experiments", id + ".json");
436
511
  const { mkdir, writeFile } = await import("node:fs/promises");
437
512
  await mkdir(path.dirname(destination), { recursive: true, mode: 0o700 });
438
- await writeFile(destination, JSON.stringify(matrix, null, 2) + "\n", { encoding: "utf8", flag: "wx", mode: 0o600 });
513
+ await writeFile(destination, JSON.stringify(matrix, null, 2) + "\n", {
514
+ encoding: "utf8",
515
+ flag: "wx",
516
+ mode: 0o600,
517
+ });
439
518
  process.stdout.write(formatModelComparisonMatrix(matrix) + "\nSaved: " + path.relative(target, destination) + "\n");
440
519
  }
441
520
  catch (error) {
@@ -449,7 +528,7 @@ async function main(args) {
449
528
  const skills = option("--skills");
450
529
  const repeats = Number(option("--repeats") ?? "2");
451
530
  const pathIndex = rest.indexOf("--path");
452
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
531
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
453
532
  if (!id || !name || !model || !skills) {
454
533
  process.stderr.write("Usage: anvil experiment-studio create --id ID --name NAME --provider openai|claude-code --model MODEL --skills skill-a,skill-b [--repeats 2] [--path PROJECT]\n");
455
534
  process.exitCode = 1;
@@ -461,16 +540,32 @@ async function main(args) {
461
540
  return;
462
541
  }
463
542
  try {
464
- const skillIds = skills.split(",").map(item => item.trim()).filter(Boolean);
465
- const matrix = createSkillExperimentMatrix({ id, name, baselineCandidateId: "baseline", treatmentCandidateId: "treatment", provider: provider, model, repeats, skillIds });
543
+ const skillIds = skills
544
+ .split(",")
545
+ .map((item) => item.trim())
546
+ .filter(Boolean);
547
+ const matrix = createSkillExperimentMatrix({
548
+ id,
549
+ name,
550
+ baselineCandidateId: "baseline",
551
+ treatmentCandidateId: "treatment",
552
+ provider: provider,
553
+ model,
554
+ repeats,
555
+ skillIds,
556
+ });
466
557
  const registry = await loadSkillRegistry(target);
467
- const unknown = skillIds.filter(skillId => !registry.some(skill => skill.id === skillId));
558
+ const unknown = skillIds.filter((skillId) => !registry.some((skill) => skill.id === skillId));
468
559
  if (unknown.length)
469
560
  throw new Error("Unknown workspace/built-in skill(s): " + unknown.join(", "));
470
561
  const destination = path.join(target, ".forge", "experiments", id + ".json");
471
562
  const { mkdir, writeFile } = await import("node:fs/promises");
472
563
  await mkdir(path.dirname(destination), { recursive: true, mode: 0o700 });
473
- await writeFile(destination, JSON.stringify(matrix, null, 2) + "\n", { encoding: "utf8", flag: "wx", mode: 0o600 });
564
+ await writeFile(destination, JSON.stringify(matrix, null, 2) + "\n", {
565
+ encoding: "utf8",
566
+ flag: "wx",
567
+ mode: 0o600,
568
+ });
474
569
  process.stdout.write(formatSkillExperimentMatrix(matrix) + "\nSaved: " + path.relative(target, destination) + "\n");
475
570
  }
476
571
  catch (error) {
@@ -482,7 +577,7 @@ async function main(args) {
482
577
  case "agent-evolution": {
483
578
  const matrixId = rest[0];
484
579
  const pathIndex = rest.indexOf("--path");
485
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
580
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
486
581
  if (!matrixId || matrixId.startsWith("--")) {
487
582
  process.stderr.write("Usage: anvil agent-evolution MATRIX_ID --path PROJECT\n");
488
583
  process.exitCode = 1;
@@ -490,7 +585,9 @@ async function main(args) {
490
585
  }
491
586
  try {
492
587
  const history = await listBenchmarkMatrices(target);
493
- const report = history.filter(item => item.matrixId === matrixId).sort((a, b) => b.createdAt.localeCompare(a.createdAt))[0];
588
+ const report = history
589
+ .filter((item) => item.matrixId === matrixId)
590
+ .sort((a, b) => b.createdAt.localeCompare(a.createdAt))[0];
494
591
  if (!report)
495
592
  throw new Error("Benchmark matrix not found: " + matrixId);
496
593
  const assessment = assessAgentEvolution(report);
@@ -528,7 +625,7 @@ async function main(args) {
528
625
  return next && !next.startsWith("--") ? next : undefined;
529
626
  };
530
627
  const pathIndex = rest.indexOf("--path");
531
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
628
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
532
629
  const rawLimit = option("--limit");
533
630
  const limit = rawLimit === undefined ? undefined : Number(rawLimit);
534
631
  if (limit !== undefined && (!Number.isSafeInteger(limit) || limit < 2 || limit > 50)) {
@@ -543,7 +640,7 @@ async function main(args) {
543
640
  const report = buildAgentEvaluationIntelligence(history, {
544
641
  ...(suiteId ? { suiteId } : {}),
545
642
  ...(candidateId ? { candidateId } : {}),
546
- ...(limit !== undefined ? { limit } : {})
643
+ ...(limit !== undefined ? { limit } : {}),
547
644
  });
548
645
  process.stdout.write(rest.includes("--json")
549
646
  ? JSON.stringify(report, null, 2) + "\n"
@@ -559,7 +656,7 @@ async function main(args) {
559
656
  }
560
657
  case "optimize-agents": {
561
658
  const pathIndex = rest.indexOf("--path");
562
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
659
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
563
660
  try {
564
661
  const matrices = await listBenchmarkMatrices(target);
565
662
  const plan = buildAgentOptimizationPlan(matrices);
@@ -575,7 +672,7 @@ async function main(args) {
575
672
  }
576
673
  case "benchmark-regression": {
577
674
  const targetIndex = rest.indexOf("--path");
578
- const target = path.resolve(targetIndex >= 0 ? rest[targetIndex + 1] ?? "." : ".");
675
+ const target = path.resolve(targetIndex >= 0 ? (rest[targetIndex + 1] ?? ".") : ".");
579
676
  const matrixId = rest.find((arg, index) => !arg.startsWith("--") && index === 0);
580
677
  try {
581
678
  const report = await loadBenchmarkRegression(target, matrixId);
@@ -591,21 +688,26 @@ async function main(args) {
591
688
  }
592
689
  case "experiment": {
593
690
  const action = rest[0] ?? "capture";
594
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
691
+ const option = (name) => {
692
+ const index = rest.indexOf(name);
693
+ return index >= 0 ? rest[index + 1] : undefined;
694
+ };
595
695
  const pathIndex = rest.indexOf("--path");
596
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
696
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
597
697
  try {
598
698
  if (action === "capture") {
599
699
  const matrixFile = option("--matrix-file");
600
700
  const suiteId = option("--suite");
601
701
  const matrixId = option("--matrix-id");
602
702
  const provider = option("--provider") ?? getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai");
603
- const model = option("--model") ?? getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ?? getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL", "gpt-4.1");
703
+ const model = option("--model") ??
704
+ getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ??
705
+ getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL", "gpt-4.1");
604
706
  if (!matrixFile || !suiteId || !matrixId)
605
707
  throw new Error("Usage: anvil experiment capture --matrix-file CONFIG.json --matrix-id ID --suite SUITE_ID [--provider ID] [--model NAME] [--path PROJECT]");
606
708
  const matrix = await loadBenchmarkMatrix(path.resolve(matrixFile));
607
709
  const suites = await loadEvaluationSuites(target);
608
- const suite = suites.find(item => item.id === suiteId);
710
+ const suite = suites.find((item) => item.id === suiteId);
609
711
  if (!suite)
610
712
  throw new Error("Evaluation suite not found: " + suiteId);
611
713
  const fingerprints = [];
@@ -613,10 +715,24 @@ async function main(args) {
613
715
  if (!candidate.configVersionId)
614
716
  continue;
615
717
  const snapshot = await loadAgentConfigVersion(target, candidate.configVersionId);
616
- fingerprints.push({ candidateId: candidate.id, configVersionId: snapshot.id, fingerprint: fingerprintAgentConfig(snapshot.files) });
718
+ fingerprints.push({
719
+ candidateId: candidate.id,
720
+ configVersionId: snapshot.id,
721
+ fingerprint: fingerprintAgentConfig(snapshot.files),
722
+ });
617
723
  }
618
724
  const baseUrl = getAnvilEnv("ANVIL_BASE_URL", "FORGE_BASE_URL");
619
- const manifest = await captureExperimentManifest({ projectPath: target, matrixFile: path.resolve(matrixFile), matrixId, suiteId, suite, provider, model, ...(baseUrl ? { baseUrl } : {}), agentConfigFingerprints: fingerprints });
725
+ const manifest = await captureExperimentManifest({
726
+ projectPath: target,
727
+ matrixFile: path.resolve(matrixFile),
728
+ matrixId,
729
+ suiteId,
730
+ suite,
731
+ provider,
732
+ model,
733
+ ...(baseUrl ? { baseUrl } : {}),
734
+ agentConfigFingerprints: fingerprints,
735
+ });
620
736
  const saved = await saveExperimentManifest(target, manifest);
621
737
  process.stdout.write(formatExperimentManifest(manifest) + "\nSaved: " + path.relative(target, saved) + "\n");
622
738
  }
@@ -630,9 +746,18 @@ async function main(args) {
630
746
  const ids = rest.slice(1).filter((arg, index, all) => !arg.startsWith("--") && all[index - 1] !== "--path");
631
747
  if (ids.length !== 2)
632
748
  throw new Error("Usage: anvil experiment compare EXPERIMENT_A EXPERIMENT_B --path PROJECT");
633
- const [a, b] = await Promise.all([loadExperimentManifest(target, ids[0]), loadExperimentManifest(target, ids[1])]);
749
+ const [a, b] = await Promise.all([
750
+ loadExperimentManifest(target, ids[0]),
751
+ loadExperimentManifest(target, ids[1]),
752
+ ]);
634
753
  const comparison = compareExperimentManifests(a, b);
635
- process.stdout.write("Experiment reproducibility: " + (comparison.compatible ? "COMPATIBLE" : "NOT COMPARABLE") + "\n" + (comparison.differences.length ? comparison.differences.map(item => " - " + item).join("\n") : " Critical manifest fields match.") + "\n");
754
+ process.stdout.write("Experiment reproducibility: " +
755
+ (comparison.compatible ? "COMPATIBLE" : "NOT COMPARABLE") +
756
+ "\n" +
757
+ (comparison.differences.length
758
+ ? comparison.differences.map((item) => " - " + item).join("\n")
759
+ : " Critical manifest fields match.") +
760
+ "\n");
636
761
  if (!comparison.compatible)
637
762
  process.exitCode = 2;
638
763
  }
@@ -646,14 +771,19 @@ async function main(args) {
646
771
  return;
647
772
  }
648
773
  case "benchmark-matrix": {
649
- const configFile = rest.find(arg => !arg.startsWith("--"));
774
+ const configFile = rest.find((arg) => !arg.startsWith("--"));
650
775
  const suiteIndex = rest.indexOf("--suite"), goalIndex = rest.indexOf("--goal"), pathIndex = rest.indexOf("--path");
651
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
776
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
652
777
  const suiteId = suiteIndex >= 0 ? rest[suiteIndex + 1] : undefined;
653
778
  const goalEnd = goalIndex >= 0 ? rest.findIndex((arg, index) => index > goalIndex && arg.startsWith("--")) : -1;
654
- const goal = goalIndex >= 0 ? rest.slice(goalIndex + 1, goalEnd >= 0 ? goalEnd : rest.length).join(" ").trim() : "";
779
+ const goal = goalIndex >= 0
780
+ ? rest
781
+ .slice(goalIndex + 1, goalEnd >= 0 ? goalEnd : rest.length)
782
+ .join(" ")
783
+ .trim()
784
+ : "";
655
785
  if (!configFile || !suiteId || !goal) {
656
- process.stderr.write("Usage: anvil benchmark-matrix CONFIG.json --suite SUITE_ID --goal \"engineering task\" [--path PROJECT]\n");
786
+ process.stderr.write('Usage: anvil benchmark-matrix CONFIG.json --suite SUITE_ID --goal "engineering task" [--path PROJECT]\n');
657
787
  process.exitCode = 1;
658
788
  return;
659
789
  }
@@ -667,7 +797,7 @@ async function main(args) {
667
797
  configVersions.set(candidate.id, { id: snapshot.id, fingerprint: fingerprintAgentConfig(snapshot.files) });
668
798
  }
669
799
  const suites = await loadEvaluationSuites(target);
670
- const suite = suites.find(item => item.id === suiteId);
800
+ const suite = suites.find((item) => item.id === suiteId);
671
801
  if (!suite)
672
802
  throw new Error("Evaluation suite not found: " + suiteId);
673
803
  const apiKey = getAnvilEnv("ANVIL_API_KEY", "FORGE_API_KEY") ?? process.env.OPENAI_API_KEY ?? "";
@@ -675,43 +805,82 @@ async function main(args) {
675
805
  const report = await executeBenchmarkMatrix(config, async (candidate, _repeat) => {
676
806
  const createProvider = (selectedModel) => candidate.provider === "claude-code"
677
807
  ? new ClaudeCodeProvider({ cwd: target, model: selectedModel })
678
- : createHostedAgentProvider({ ...process.env, ANVIL_AI_PROVIDER: candidate.provider, FORGE_AI_PROVIDER: candidate.provider, ...(apiKey ? { ANVIL_API_KEY: apiKey, FORGE_API_KEY: apiKey } : {}), ...(baseUrl ? { ANVIL_BASE_URL: baseUrl, FORGE_BASE_URL: baseUrl } : {}) }, selectedModel);
808
+ : createHostedAgentProvider({
809
+ ...process.env,
810
+ ANVIL_AI_PROVIDER: candidate.provider,
811
+ FORGE_AI_PROVIDER: candidate.provider,
812
+ ...(apiKey ? { ANVIL_API_KEY: apiKey, FORGE_API_KEY: apiKey } : {}),
813
+ ...(baseUrl ? { ANVIL_BASE_URL: baseUrl, FORGE_BASE_URL: baseUrl } : {}),
814
+ }, selectedModel);
679
815
  const router = new ModelRouter({
680
- fastModel: candidate.model, balancedModel: candidate.model, reasoningModel: candidate.model,
816
+ fastModel: candidate.model,
817
+ balancedModel: candidate.model,
818
+ reasoningModel: candidate.model,
681
819
  retryAttempts: Number(getAnvilEnv("ANVIL_MODEL_RETRY_ATTEMPTS", "FORGE_MODEL_RETRY_ATTEMPTS", "2")),
682
820
  retryDelayMs: Number(getAnvilEnv("ANVIL_MODEL_RETRY_DELAY_MS", "FORGE_MODEL_RETRY_DELAY_MS", "500")),
683
- ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK") ? { maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")) } : {}),
684
- ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK") ? { maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")) } : {}),
685
- ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN") ? { maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")) } : {}),
686
- ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT") ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() } : {}),
687
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS") ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) } : {}),
688
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT") ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() } : {}),
821
+ ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")
822
+ ? {
823
+ maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")),
824
+ }
825
+ : {}),
826
+ ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")
827
+ ? {
828
+ maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")),
829
+ }
830
+ : {}),
831
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")
832
+ ? {
833
+ maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")),
834
+ }
835
+ : {}),
836
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT")
837
+ ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() }
838
+ : {}),
839
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")
840
+ ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) }
841
+ : {}),
842
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT")
843
+ ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() }
844
+ : {}),
689
845
  createProvider,
690
846
  });
691
847
  const configWorkspace = await mkdtemp(path.join(os.tmpdir(), "forge-candidate-config-"));
692
848
  try {
693
849
  const snapshotId = candidate.configVersionId;
694
- const snapshot = snapshotId ? await materializeAgentConfigVersion(target, snapshotId, configWorkspace) : undefined;
695
- if (snapshotId && !snapshot?.files.some(file => file.path === "hooks.json")) {
696
- throw new Error("Configuration snapshot " + snapshotId + " predates policy-hook isolation. Capture a new snapshot before benchmarking this candidate.");
850
+ const snapshot = snapshotId
851
+ ? await materializeAgentConfigVersion(target, snapshotId, configWorkspace)
852
+ : undefined;
853
+ if (snapshotId && !snapshot?.files.some((file) => file.path === "hooks.json")) {
854
+ throw new Error("Configuration snapshot " +
855
+ snapshotId +
856
+ " predates policy-hook isolation. Capture a new snapshot before benchmarking this candidate.");
697
857
  }
698
858
  const configRoot = snapshotId ? configWorkspace : target;
699
859
  const run = createReviewPlan(goal);
700
860
  const skillRegistry = await loadSkillRegistry(configRoot);
701
861
  const agentRegistry = await loadAgentRegistry(configRoot);
702
862
  const requestedSkillIds = candidate.skillIds;
703
- const unknownSkills = (requestedSkillIds ?? []).filter(id => !skillRegistry.some(skill => skill.id === id));
863
+ const unknownSkills = (requestedSkillIds ?? []).filter((id) => !skillRegistry.some((skill) => skill.id === id));
704
864
  if (unknownSkills.length)
705
865
  throw new Error("Candidate " + candidate.id + " references unknown skill(s): " + unknownSkills.join(", "));
706
- const selectedSkillsForTask = (task) => requestedSkillIds === undefined ? selectSkills(skillRegistry, task.agent + " " + task.title + " " + task.prompt) : skillRegistry.filter(skill => requestedSkillIds.includes(skill.id));
707
- const agentGuidanceForTask = (task) => agentRegistry.find(agent => agent.id === task.agent)?.instructions ?? "";
866
+ const selectedSkillsForTask = (task) => requestedSkillIds === undefined
867
+ ? selectSkills(skillRegistry, task.agent + " " + task.title + " " + task.prompt)
868
+ : skillRegistry.filter((skill) => requestedSkillIds.includes(skill.id));
869
+ const agentGuidanceForTask = (task) => agentRegistry.find((agent) => agent.id === task.agent)?.instructions ?? "";
708
870
  const permissions = await loadPermissionPolicy(configRoot);
709
871
  if (permissions.mode !== "proposal-only" || permissions.allowedTools.length > 0)
710
- throw new Error("Benchmark candidate " + candidate.id + " requests unsupported tool permissions. FORGE currently enforces proposal-only execution with no enabled tools.");
872
+ throw new Error("Benchmark candidate " +
873
+ candidate.id +
874
+ " requests unsupported tool permissions. FORGE currently enforces proposal-only execution with no enabled tools.");
711
875
  const hooks = await loadHooks(configRoot);
712
876
  for (const record of run.tasks) {
713
877
  const applied = applyBeforeAgentHooks(record.task, hooks);
714
- record.task = { ...applied.task, prompt: applied.context.length ? applied.task.prompt + "\n\nHook guidance:\n" + applied.context.join("\n") : applied.task.prompt };
878
+ record.task = {
879
+ ...applied.task,
880
+ prompt: applied.context.length
881
+ ? applied.task.prompt + "\n\nHook guidance:\n" + applied.context.join("\n")
882
+ : applied.task.prompt,
883
+ };
715
884
  if (applied.blockedReason) {
716
885
  record.status = "blocked";
717
886
  record.error = "Blocked by policy hook: " + applied.blockedReason;
@@ -720,17 +889,49 @@ async function main(args) {
720
889
  const memory = await buildProjectMemoryContext(target, goal);
721
890
  const projectContext = memory.text;
722
891
  const started = Date.now();
723
- const completed = await advanceRun(run, router, new Set(), { maxConcurrency: 4, projectContext, projectContextForTask: async (task) => [agentGuidanceForTask(task), formatSkillContext(selectedSkillsForTask(task)), await buildAgentWorkspaceContext(target, task)].filter(Boolean).join("\n\n") });
892
+ const completed = await advanceRun(run, router, new Set(), {
893
+ maxConcurrency: 4,
894
+ projectContext,
895
+ projectContextForTask: async (task) => [
896
+ agentGuidanceForTask(task),
897
+ formatSkillContext(selectedSkillsForTask(task)),
898
+ await buildAgentWorkspaceContext(target, task),
899
+ ]
900
+ .filter(Boolean)
901
+ .join("\n\n"),
902
+ });
724
903
  const durationMs = Date.now() - started;
725
904
  await saveRun(target, completed);
726
905
  await saveExecutionTrace(target, completed.id, router.traces);
727
906
  const evaluation = evaluateRun(completed, suite);
728
907
  await saveEvaluationReport(target, evaluation);
729
908
  return {
730
- runId: completed.id, score: evaluation.score, passed: evaluation.passed,
909
+ runId: completed.id,
910
+ score: evaluation.score,
911
+ passed: evaluation.passed,
731
912
  durationMs,
732
913
  estimatedTokens: router.traces.reduce((sum, item) => sum + item.inputTokensEstimate + item.outputTokensEstimate, 0),
733
- agentScores: evaluation.cases.filter(item => item.status === "scored").map(item => { const task = completed.tasks.find(record => record.task.id === item.taskId)?.task; const appliedSkills = task ? selectedSkillsForTask(task) : []; return { agent: item.agent, caseId: item.caseId, score: item.score, passed: item.passed, skillIds: appliedSkills.map(skill => skill.id), failedCriteria: item.criteria.filter(criterion => !criterion.passed).map(criterion => ({ id: criterion.id, failedChecks: criterion.checks.filter(check => !check.passed).map(check => check.rule + ": " + check.detail) })) }; }),
914
+ agentScores: evaluation.cases
915
+ .filter((item) => item.status === "scored")
916
+ .map((item) => {
917
+ const task = completed.tasks.find((record) => record.task.id === item.taskId)?.task;
918
+ const appliedSkills = task ? selectedSkillsForTask(task) : [];
919
+ return {
920
+ agent: item.agent,
921
+ caseId: item.caseId,
922
+ score: item.score,
923
+ passed: item.passed,
924
+ skillIds: appliedSkills.map((skill) => skill.id),
925
+ failedCriteria: item.criteria
926
+ .filter((criterion) => !criterion.passed)
927
+ .map((criterion) => ({
928
+ id: criterion.id,
929
+ failedChecks: criterion.checks
930
+ .filter((check) => !check.passed)
931
+ .map((check) => check.rule + ": " + check.detail),
932
+ })),
933
+ };
934
+ }),
734
935
  };
735
936
  }
736
937
  finally {
@@ -740,19 +941,34 @@ async function main(args) {
740
941
  report.suiteId = suite.id;
741
942
  report.goal = goal;
742
943
  if (configVersions.size)
743
- report.candidateProvenance = [...configVersions.entries()].map(([candidateId, item]) => ({ candidateId, configVersionId: item.id, fingerprint: item.fingerprint }));
944
+ report.candidateProvenance = [...configVersions.entries()].map(([candidateId, item]) => ({
945
+ candidateId,
946
+ configVersionId: item.id,
947
+ fingerprint: item.fingerprint,
948
+ }));
744
949
  const experimentManifest = await captureExperimentManifest({
745
- projectPath: target, matrixFile: path.resolve(configFile), matrixId: config.id, suiteId: suite.id, suite,
746
- provider: "matrix", model: "candidate-specific",
950
+ projectPath: target,
951
+ matrixFile: path.resolve(configFile),
952
+ matrixId: config.id,
953
+ suiteId: suite.id,
954
+ suite,
955
+ provider: "matrix",
956
+ model: "candidate-specific",
747
957
  ...(process.env.FORGE_BASE_URL ? { baseUrl: process.env.FORGE_BASE_URL } : {}),
748
- agentConfigFingerprints: report.candidateProvenance ?? []
958
+ agentConfigFingerprints: report.candidateProvenance ?? [],
749
959
  });
750
960
  const manifestPath = await saveExperimentManifest(target, experimentManifest);
751
961
  report.experimentManifest = { id: experimentManifest.id, fingerprint: experimentManifest.fingerprint };
752
962
  const saved = await saveBenchmarkMatrix(target, report);
753
- process.stdout.write("Experiment manifest: " + experimentManifest.id + " · sha256:" + experimentManifest.fingerprint + "\nManifest saved: " + path.relative(target, manifestPath) + "\n");
963
+ process.stdout.write("Experiment manifest: " +
964
+ experimentManifest.id +
965
+ " · sha256:" +
966
+ experimentManifest.fingerprint +
967
+ "\nManifest saved: " +
968
+ path.relative(target, manifestPath) +
969
+ "\n");
754
970
  process.stdout.write(formatBenchmarkMatrix(report) + "\nSaved: " + path.relative(target, saved) + "\n");
755
- if (report.results.some(item => item.status === "failed") || report.results.some(item => !item.passed))
971
+ if (report.results.some((item) => item.status === "failed") || report.results.some((item) => !item.passed))
756
972
  process.exitCode = 2;
757
973
  }
758
974
  catch (error) {
@@ -764,12 +980,14 @@ async function main(args) {
764
980
  case "benchmark": {
765
981
  const pathIndex = rest.indexOf("--path"), suiteIndex = rest.indexOf("--suite");
766
982
  const runId = rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path" && rest[index - 1] !== "--suite");
767
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
983
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
768
984
  const suiteId = suiteIndex >= 0 ? rest[suiteIndex + 1] : undefined;
769
985
  try {
770
986
  if (!runId) {
771
987
  const documents = await listBenchmarks(target);
772
- process.stdout.write(documents.length ? documents.map(document => formatBenchmark(document)).join("\n\n") + "\n" : "No benchmark history. Evaluate and record a run with: forge benchmark RUN_ID --path .\n");
988
+ process.stdout.write(documents.length
989
+ ? documents.map((document) => formatBenchmark(document)).join("\n\n") + "\n"
990
+ : "No benchmark history. Evaluate and record a run with: forge benchmark RUN_ID --path .\n");
773
991
  return;
774
992
  }
775
993
  const report = await loadEvaluationReport(target, runId, suiteId);
@@ -784,9 +1002,9 @@ async function main(args) {
784
1002
  return;
785
1003
  }
786
1004
  case "evaluate": {
787
- const runId = rest.find(arg => !arg.startsWith("--"));
1005
+ const runId = rest.find((arg) => !arg.startsWith("--"));
788
1006
  const pathIndex = rest.indexOf("--path"), suiteIndex = rest.indexOf("--suite");
789
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1007
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
790
1008
  const suiteId = suiteIndex >= 0 ? rest[suiteIndex + 1] : undefined;
791
1009
  if (!runId) {
792
1010
  process.stderr.write("Usage: anvil evaluate RUNID [--suite SUITE_ID] [--path PROJECT]\n");
@@ -795,20 +1013,27 @@ async function main(args) {
795
1013
  }
796
1014
  try {
797
1015
  const suites = await loadEvaluationSuites(target);
798
- const suite = suiteId ? suites.find(item => item.id === suiteId) : suites[0];
1016
+ const suite = suiteId ? suites.find((item) => item.id === suiteId) : suites[0];
799
1017
  if (!suite)
800
- throw new Error(suiteId ? "Evaluation suite not found: " + suiteId : "No evaluation suite found. Add one under .forge/evals/.");
1018
+ throw new Error(suiteId
1019
+ ? "Evaluation suite not found: " + suiteId
1020
+ : "No evaluation suite found. Add one under .forge/evals/.");
801
1021
  const run = await loadRun(target, runId);
802
1022
  const judgeEnabled = rest.includes("--judge");
803
1023
  let report;
804
1024
  if (judgeEnabled) {
805
1025
  const providerChoice = getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai");
806
- const model = getAnvilEnv("ANVIL_EVAL_JUDGE_MODEL", "FORGE_EVAL_JUDGE_MODEL") ?? getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ?? getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL") ?? (providerChoice === "claude-code" ? "sonnet" : "gpt-4.1");
1026
+ const model = getAnvilEnv("ANVIL_EVAL_JUDGE_MODEL", "FORGE_EVAL_JUDGE_MODEL") ??
1027
+ getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ??
1028
+ getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL") ??
1029
+ (providerChoice === "claude-code" ? "sonnet" : "gpt-4.1");
807
1030
  const apiKey = getAnvilEnv("ANVIL_API_KEY", "FORGE_API_KEY") ?? process.env.OPENAI_API_KEY ?? "";
808
1031
  const baseUrl = getAnvilEnv("ANVIL_BASE_URL", "FORGE_BASE_URL", "https://api.openai.com/v1");
809
1032
  if (providerChoice !== "claude-code" && !apiKey)
810
1033
  throw new Error("Set ANVIL_API_KEY, FORGE_API_KEY, or OPENAI_API_KEY to use model-judged evaluation.");
811
- const judgeProvider = providerChoice === "claude-code" ? new ClaudeCodeProvider({ cwd: target, model }) : new OpenAICompatibleProvider({ apiKey, model, baseUrl });
1034
+ const judgeProvider = providerChoice === "claude-code"
1035
+ ? new ClaudeCodeProvider({ cwd: target, model })
1036
+ : new OpenAICompatibleProvider({ apiKey, model, baseUrl });
812
1037
  report = await evaluateRunWithJudge(run, suite, judgeProvider);
813
1038
  }
814
1039
  else
@@ -816,7 +1041,12 @@ async function main(args) {
816
1041
  const reportPath = await saveEvaluationReport(target, report);
817
1042
  process.stdout.write(formatEvaluationReport(report) + "\n");
818
1043
  if ("judgedCases" in report) {
819
- process.stdout.write("Model judge: " + report.judge.provider + (report.judge.model ? " / " + report.judge.model : "") + " · " + report.judge.evaluatedCases + " cases\n");
1044
+ process.stdout.write("Model judge: " +
1045
+ report.judge.provider +
1046
+ (report.judge.model ? " / " + report.judge.model : "") +
1047
+ " · " +
1048
+ report.judge.evaluatedCases +
1049
+ " cases\n");
820
1050
  for (const item of report.judgedCases)
821
1051
  process.stdout.write(" " + item.caseId + " · " + item.result.score + "/100 · " + item.result.rationale + "\n");
822
1052
  }
@@ -831,7 +1061,7 @@ async function main(args) {
831
1061
  return;
832
1062
  }
833
1063
  case "agents": {
834
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
1064
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
835
1065
  try {
836
1066
  const agents = await loadAgentRegistry(path.resolve(target));
837
1067
  process.stdout.write("FORGE agent registry\n\n");
@@ -847,22 +1077,28 @@ async function main(args) {
847
1077
  return;
848
1078
  }
849
1079
  case "skills": {
850
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
851
- const target = path.resolve(rest.find(arg => !arg.startsWith("--") && arg !== option("--show") && arg !== option("--match")) ?? ".");
1080
+ const option = (name) => {
1081
+ const index = rest.indexOf(name);
1082
+ return index >= 0 ? rest[index + 1] : undefined;
1083
+ };
1084
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--") && arg !== option("--show") && arg !== option("--match")) ?? ".");
852
1085
  try {
853
1086
  const skills = await loadSkillRegistry(target);
854
1087
  const showId = option("--show");
855
1088
  const matchGoal = option("--match");
856
1089
  if (showId) {
857
- const skill = skills.find(item => item.id === showId);
1090
+ const skill = skills.find((item) => item.id === showId);
858
1091
  if (!skill)
859
1092
  throw new Error("Skill not found: " + showId);
860
1093
  process.stdout.write((rest.includes("--json") ? JSON.stringify(skill, null, 2) : formatSkillContext([skill])) + "\n");
861
1094
  }
862
1095
  else if (matchGoal !== undefined) {
863
1096
  const selected = selectSkills(skills, matchGoal);
864
- process.stdout.write(rest.includes("--json") ? JSON.stringify(selected, null, 2) + "\n" :
865
- selected.length ? selected.map(skill => skill.id + " " + skill.name + " — " + skill.description).join("\n") + "\n" : "No skills matched that goal.\n");
1097
+ process.stdout.write(rest.includes("--json")
1098
+ ? JSON.stringify(selected, null, 2) + "\n"
1099
+ : selected.length
1100
+ ? selected.map((skill) => skill.id + " " + skill.name + " — " + skill.description).join("\n") + "\n"
1101
+ : "No skills matched that goal.\n");
866
1102
  }
867
1103
  else if (rest.includes("--json")) {
868
1104
  process.stdout.write(JSON.stringify(skills, null, 2) + "\n");
@@ -871,7 +1107,7 @@ async function main(args) {
871
1107
  process.stdout.write("FORGE skill registry (" + skills.length + ")\n\n");
872
1108
  for (const skill of skills)
873
1109
  process.stdout.write(" " + skill.id.padEnd(24) + " " + skill.name + " [" + skill.source + "]\n " + skill.description + "\n");
874
- process.stdout.write("\nUse --show ID to inspect instructions or --match \"GOAL\" to preview task selection.\n");
1110
+ process.stdout.write('\nUse --show ID to inspect instructions or --match "GOAL" to preview task selection.\n');
875
1111
  }
876
1112
  }
877
1113
  catch (error) {
@@ -881,12 +1117,10 @@ async function main(args) {
881
1117
  return;
882
1118
  }
883
1119
  case "agent-security": {
884
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
1120
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
885
1121
  try {
886
1122
  const report = await auditAgentConfiguration(target);
887
- process.stdout.write(rest.includes("--json")
888
- ? JSON.stringify(report, null, 2) + "\n"
889
- : formatAgentSecurityReport(report) + "\n");
1123
+ process.stdout.write(rest.includes("--json") ? JSON.stringify(report, null, 2) + "\n" : formatAgentSecurityReport(report) + "\n");
890
1124
  if (report.summary.critical > 0)
891
1125
  process.exitCode = 2;
892
1126
  }
@@ -899,7 +1133,7 @@ async function main(args) {
899
1133
  case "hook":
900
1134
  case "hooks": {
901
1135
  if (rest[0] === "install" || rest.includes("--install")) {
902
- const target = path.resolve(rest.find(arg => !arg.startsWith("--") && arg !== "install" && arg !== "pre-commit") ?? ".");
1136
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--") && arg !== "install" && arg !== "pre-commit") ?? ".");
903
1137
  try {
904
1138
  const { installPrecommitHook } = await import("./core/project-cleaner.js");
905
1139
  const result = await installPrecommitHook(target);
@@ -913,10 +1147,13 @@ async function main(args) {
913
1147
  }
914
1148
  return;
915
1149
  }
916
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
1150
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
917
1151
  try {
918
1152
  const hooks = await loadHooks(target);
919
- process.stdout.write(hooks.length ? hooks.map(hook => hook.id + " [" + hook.event + "] " + hook.action + ": " + hook.message).join("\n") + "\n" : "No lifecycle hooks configured.\n");
1153
+ process.stdout.write(hooks.length
1154
+ ? hooks.map((hook) => hook.id + " [" + hook.event + "] " + hook.action + ": " + hook.message).join("\n") +
1155
+ "\n"
1156
+ : "No lifecycle hooks configured.\n");
920
1157
  }
921
1158
  catch (error) {
922
1159
  process.stderr.write("Could not load hooks: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -926,19 +1163,24 @@ async function main(args) {
926
1163
  }
927
1164
  case "memory": {
928
1165
  const [action = "list", ...memoryArgs] = rest;
929
- const option = (name) => { const index = memoryArgs.indexOf(name); return index >= 0 ? memoryArgs[index + 1] : undefined; };
1166
+ const option = (name) => {
1167
+ const index = memoryArgs.indexOf(name);
1168
+ return index >= 0 ? memoryArgs[index + 1] : undefined;
1169
+ };
930
1170
  const skipValues = new Set();
931
1171
  for (let i = 0; i < memoryArgs.length; i++) {
932
1172
  if (memoryArgs[i]?.startsWith("--"))
933
1173
  skipValues.add(i + 1);
934
1174
  }
935
1175
  const positional = memoryArgs.filter((arg, i) => !arg.startsWith("--") && !skipValues.has(i));
936
- const targetPath = positional.find(arg => !arg.startsWith("mem-")) ?? ".";
1176
+ const targetPath = positional.find((arg) => !arg.startsWith("mem-")) ?? ".";
937
1177
  const target = path.resolve(targetPath);
938
1178
  try {
939
1179
  if (action === "list") {
940
1180
  const entries = await listProjectMemory(target);
941
- process.stdout.write(entries.length ? entries.map(entry => `${entry.id} [${entry.kind}] ${entry.content}`).join("\n") + "\n" : "No project memory entries.\n");
1181
+ process.stdout.write(entries.length
1182
+ ? entries.map((entry) => `${entry.id} [${entry.kind}] ${entry.content}`).join("\n") + "\n"
1183
+ : "No project memory entries.\n");
942
1184
  }
943
1185
  else if (action === "add") {
944
1186
  const kind = option("--kind");
@@ -951,7 +1193,7 @@ async function main(args) {
951
1193
  process.stdout.write(`Saved ${entry.id} to project memory\n`);
952
1194
  }
953
1195
  else if (action === "remove") {
954
- const id = memoryArgs.find(arg => arg.startsWith("mem-"));
1196
+ const id = memoryArgs.find((arg) => arg.startsWith("mem-"));
955
1197
  if (!id)
956
1198
  throw new Error("Usage: anvil memory remove [path] mem-ID");
957
1199
  const removed = await removeProjectMemory(target, id);
@@ -974,7 +1216,10 @@ async function main(args) {
974
1216
  }
975
1217
  case "mcp": {
976
1218
  const [subcommand = "info", ...mcpArgs] = rest;
977
- const option = (name) => { const index = mcpArgs.indexOf(name); return index >= 0 ? mcpArgs[index + 1] : undefined; };
1219
+ const option = (name) => {
1220
+ const index = mcpArgs.indexOf(name);
1221
+ return index >= 0 ? mcpArgs[index + 1] : undefined;
1222
+ };
978
1223
  const skipValues = new Set();
979
1224
  for (let i = 0; i < mcpArgs.length; i++) {
980
1225
  if (mcpArgs[i]?.startsWith("--"))
@@ -1028,7 +1273,10 @@ Subcommands:
1028
1273
  return;
1029
1274
  }
1030
1275
  const [action = "list", ...adrArgs] = rest;
1031
- const option = (name) => { const index = adrArgs.indexOf(name); return index >= 0 ? adrArgs[index + 1] : undefined; };
1276
+ const option = (name) => {
1277
+ const index = adrArgs.indexOf(name);
1278
+ return index >= 0 ? adrArgs[index + 1] : undefined;
1279
+ };
1032
1280
  const skipValues = new Set();
1033
1281
  for (let i = 0; i < adrArgs.length; i++) {
1034
1282
  if (adrArgs[i]?.startsWith("--"))
@@ -1039,12 +1287,14 @@ Subcommands:
1039
1287
  if (action === "init") {
1040
1288
  const target = path.resolve(positional[0] ?? ".");
1041
1289
  const res = await initAdrDirectory(target);
1042
- process.stdout.write(res.created ? `Initialized Architecture Decision Records in ${res.path}\n` : `ADR directory already initialized in ${res.path}\n`);
1290
+ process.stdout.write(res.created
1291
+ ? `Initialized Architecture Decision Records in ${res.path}\n`
1292
+ : `ADR directory already initialized in ${res.path}\n`);
1043
1293
  }
1044
1294
  else if (action === "new") {
1045
1295
  const title = positional[0];
1046
1296
  if (!title)
1047
- throw new Error("Usage: anvil adr new \"Title of decision\" [path] [--status Proposed|Accepted]");
1297
+ throw new Error('Usage: anvil adr new "Title of decision" [path] [--status Proposed|Accepted]');
1048
1298
  const target = path.resolve(positional[1] ?? ".");
1049
1299
  const status = (option("--status") ?? "Accepted");
1050
1300
  const context = option("--context");
@@ -1056,7 +1306,9 @@ Subcommands:
1056
1306
  const target = path.resolve(positional[0] ?? ".");
1057
1307
  const res = await syncMemoryToAdrs(target);
1058
1308
  process.stdout.write(res.syncedCount > 0
1059
- ? `Synced ${res.syncedCount} decisions from project memory to docs/adr/:\n` + res.createdAdrs.map(a => ` - [${a.id}] ${a.filename}`).join("\n") + "\n"
1309
+ ? `Synced ${res.syncedCount} decisions from project memory to docs/adr/:\n` +
1310
+ res.createdAdrs.map((a) => ` - [${a.id}] ${a.filename}`).join("\n") +
1311
+ "\n"
1060
1312
  : `All project memory decisions are already synced to docs/adr/.\n`);
1061
1313
  }
1062
1314
  else if (action === "list") {
@@ -1097,14 +1349,16 @@ Options:
1097
1349
  }
1098
1350
  const isJson = rest.includes("--json");
1099
1351
  const shouldFix = rest.includes("--fix");
1100
- const positional = rest.filter(arg => !arg.startsWith("--"));
1352
+ const positional = rest.filter((arg) => !arg.startsWith("--"));
1101
1353
  const target = path.resolve(positional[0] ?? ".");
1102
1354
  try {
1103
1355
  const report = await scanProjectSecurity(target);
1104
1356
  if (shouldFix && report.findings.length > 0) {
1105
1357
  const fixResult = await applySecurityFixes(target, report.findings);
1106
1358
  if (fixResult.fixedCount > 0) {
1107
- process.stdout.write(`Applied ${fixResult.fixedCount} security fix(es):\n` + fixResult.details.map(d => ` ✓ ${d}`).join("\n") + "\n\n");
1359
+ process.stdout.write(`Applied ${fixResult.fixedCount} security fix(es):\n` +
1360
+ fixResult.details.map((d) => ` ✓ ${d}`).join("\n") +
1361
+ "\n\n");
1108
1362
  }
1109
1363
  }
1110
1364
  if (isJson) {
@@ -1132,11 +1386,13 @@ and code quality rules against policies. Emits human-readable reports, machine-r
1132
1386
  or OASIS SARIF v2.1.0 for PR comments and GitHub Security tab integration.
1133
1387
 
1134
1388
  Options:
1135
- --sarif PATH Export findings to an OASIS SARIF v2.1.0 report file
1136
- --policy PATH Custom JSON policy file (defaults to .anvil/policy.json if present)
1137
- --fail-on-warnings Fail gate (exit 1) on medium-severity warnings as well as critical/high errors
1138
- --json Output full gate evaluation in machine-readable JSON format
1139
- --help Show this help message
1389
+ --sarif PATH Export findings to an OASIS SARIF v2.1.0 report file
1390
+ --policy PATH Custom JSON policy file (defaults to .anvil/policy.json if present)
1391
+ --baseline PATH Baseline JSON file to suppress known existing findings
1392
+ --generate-baseline PATH Export current findings as a baseline JSON file
1393
+ --fail-on-warnings Fail gate (exit 1) on medium-severity warnings as well as critical/high errors
1394
+ --json Output full gate evaluation in machine-readable JSON format
1395
+ --help Show this help message
1140
1396
 
1141
1397
  Exit codes:
1142
1398
  0 Gate passed: all policy thresholds and security checks satisfied.
@@ -1155,13 +1411,35 @@ Exit codes:
1155
1411
  const failOnWarnings = rest.includes("--fail-on-warnings");
1156
1412
  const sarifPath = option("--sarif");
1157
1413
  const policyPath = option("--policy");
1158
- const positional = rest.filter(arg => !arg.startsWith("--") && arg !== sarifPath && arg !== policyPath);
1414
+ const baselinePath = option("--baseline");
1415
+ const generateBaselinePath = option("--generate-baseline");
1416
+ const positional = rest.filter((arg) => !arg.startsWith("--") &&
1417
+ arg !== sarifPath &&
1418
+ arg !== policyPath &&
1419
+ arg !== baselinePath &&
1420
+ arg !== generateBaselinePath);
1159
1421
  const target = path.resolve(positional[0] ?? ".");
1160
1422
  try {
1161
- const securityReport = await scanProjectSecurity(target);
1162
- const reviewReport = await reviewProject(target, { all: true });
1423
+ const evaluation = await evaluateGate(target, {
1424
+ policyPath,
1425
+ baselinePath,
1426
+ failOnWarnings,
1427
+ });
1428
+ if (generateBaselinePath) {
1429
+ const outResolved = path.resolve(generateBaselinePath);
1430
+ await mkdir(path.dirname(outResolved), { recursive: true });
1431
+ await writeFile(outResolved, generateBaselineJson(evaluation.activeFindings), "utf8");
1432
+ if (!isJson) {
1433
+ process.stdout.write(`Generated baseline with ${evaluation.activeFindings.length} findings to ${path.relative(process.cwd(), outResolved)}\n`);
1434
+ }
1435
+ }
1163
1436
  if (sarifPath) {
1164
- const sarifOutput = generateSarif({ securityReport, reviewReport, targetRoot: target });
1437
+ const sarifOutput = generateSarif({
1438
+ activeFindings: evaluation.activeFindings,
1439
+ suppressedFindings: evaluation.suppressedFindings,
1440
+ filterSuppressed: true,
1441
+ targetRoot: target,
1442
+ });
1165
1443
  const outResolved = path.resolve(sarifPath);
1166
1444
  await mkdir(path.dirname(outResolved), { recursive: true });
1167
1445
  await writeFile(outResolved, sarifOutput, "utf8");
@@ -1169,42 +1447,46 @@ Exit codes:
1169
1447
  process.stdout.write(`Exported SARIF report to ${path.relative(process.cwd(), outResolved)}\n`);
1170
1448
  }
1171
1449
  }
1172
- const criticalCount = securityReport.summary.critical;
1173
- const highCount = securityReport.summary.high + reviewReport.summary.high;
1174
- const mediumCount = securityReport.summary.medium + reviewReport.summary.medium;
1175
- const passed = criticalCount === 0 && highCount === 0 && (!failOnWarnings || mediumCount === 0);
1176
1450
  if (isJson) {
1177
1451
  process.stdout.write(JSON.stringify({
1178
- gate: passed ? "PASSED" : "FAILED",
1452
+ gate: evaluation.passed ? "PASSED" : "FAILED",
1179
1453
  target,
1180
- summary: {
1181
- critical: criticalCount,
1182
- high: highCount,
1183
- medium: mediumCount,
1184
- low: securityReport.summary.low + reviewReport.summary.low
1185
- },
1186
- security: securityReport,
1187
- review: reviewReport
1454
+ policySource: evaluation.policySource,
1455
+ baselineSource: evaluation.baselineSource,
1456
+ violations: evaluation.violations,
1457
+ summary: evaluation.summary,
1458
+ activeFindings: evaluation.activeFindings,
1459
+ suppressedFindings: evaluation.suppressedFindings,
1188
1460
  }, null, 2) + "\n");
1189
1461
  }
1190
1462
  else {
1191
- process.stdout.write([
1463
+ const lines = [
1192
1464
  `# ANVIL CI Quality & Security Gate`,
1193
1465
  `Target: ${target}`,
1194
- `Status: ${passed ? "PASSED" : "FAILED"}`,
1195
- ``,
1196
- `Findings Summary:`,
1197
- ` Critical: ${criticalCount}`,
1198
- ` High: ${highCount}`,
1199
- ` Medium: ${mediumCount}`,
1200
- ` Low: ${securityReport.summary.low + reviewReport.summary.low}`,
1201
- ``,
1202
- passed
1203
- ? `✓ Gate passed: 0 critical/high findings detected.`
1204
- : `✗ Gate failed: ${criticalCount + highCount} critical or high findings require remediation before merging.`
1205
- ].join("\n") + "\n");
1466
+ `Status: ${evaluation.passed ? "PASSED" : "FAILED"}`,
1467
+ `Policy: ${evaluation.policySource}`,
1468
+ ];
1469
+ if (evaluation.baselineSource) {
1470
+ lines.push(`Baseline: ${evaluation.baselineSource} (${evaluation.summary.suppressedByBaseline} findings suppressed)`);
1471
+ }
1472
+ lines.push(``, `Findings Summary:`, ` Critical: ${evaluation.summary.critical}`, ` High: ${evaluation.summary.high}`, ` Medium: ${evaluation.summary.medium}`, ` Low: ${evaluation.summary.low}`);
1473
+ if (evaluation.summary.suppressedByBaseline > 0) {
1474
+ lines.push(` Suppressed by baseline: ${evaluation.summary.suppressedByBaseline}`);
1475
+ }
1476
+ lines.push(``);
1477
+ if (evaluation.violations.length > 0) {
1478
+ lines.push(`Policy Violations:`);
1479
+ for (const v of evaluation.violations) {
1480
+ lines.push(` ✗ ${v}`);
1481
+ }
1482
+ lines.push(``);
1483
+ }
1484
+ lines.push(evaluation.passed
1485
+ ? `✓ Gate passed: all policy thresholds satisfied.`
1486
+ : `✗ Gate failed: ${evaluation.violations.length} policy violations detected.`);
1487
+ process.stdout.write(lines.join("\n") + "\n");
1206
1488
  }
1207
- if (!passed) {
1489
+ if (!evaluation.passed) {
1208
1490
  process.exitCode = 1;
1209
1491
  }
1210
1492
  }
@@ -1218,8 +1500,13 @@ Exit codes:
1218
1500
  const goalIndex = rest.indexOf("--goal");
1219
1501
  const pathIndex = rest.indexOf("--path");
1220
1502
  const goalEnd = pathIndex >= 0 ? pathIndex : rest.length;
1221
- const goal = goalIndex >= 0 ? rest.slice(goalIndex + 1, goalEnd).join(" ").trim() : "";
1222
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1503
+ const goal = goalIndex >= 0
1504
+ ? rest
1505
+ .slice(goalIndex + 1, goalEnd)
1506
+ .join(" ")
1507
+ .trim()
1508
+ : "";
1509
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1223
1510
  try {
1224
1511
  const plan = createReviewPlan(goal);
1225
1512
  const saved = await saveRun(target, plan);
@@ -1236,7 +1523,7 @@ Exit codes:
1236
1523
  case "replay": {
1237
1524
  const runId = rest[0] ?? "";
1238
1525
  const pathIndex = rest.indexOf("--path");
1239
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1526
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1240
1527
  try {
1241
1528
  if (!runId)
1242
1529
  throw new Error("Usage: anvil replay RUN_ID [--path PROJECT]");
@@ -1257,13 +1544,13 @@ Exit codes:
1257
1544
  const runId = rest[0] ?? "";
1258
1545
  const taskId = rest[1] ?? "";
1259
1546
  const pathIndex = rest.indexOf("--path");
1260
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1547
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1261
1548
  const reviewerIndex = rest.indexOf("--reviewer");
1262
1549
  const reasonIndex = rest.indexOf("--reason");
1263
1550
  try {
1264
1551
  if (!runId || !taskId)
1265
1552
  throw new Error("Usage: anvil approve RUN_ID TASK_ID [--reviewer NAME] [--reason TEXT] [--path PROJECT]");
1266
- const updated = await decideTaskApproval(target, runId, taskId, "approved", reviewerIndex >= 0 ? rest[reviewerIndex + 1] ?? "local-user" : "local-user", reasonIndex >= 0 ? rest.slice(reasonIndex + 1, pathIndex >= 0 ? pathIndex : rest.length).join(" ") : undefined);
1553
+ const updated = await decideTaskApproval(target, runId, taskId, "approved", reviewerIndex >= 0 ? (rest[reviewerIndex + 1] ?? "local-user") : "local-user", reasonIndex >= 0 ? rest.slice(reasonIndex + 1, pathIndex >= 0 ? pathIndex : rest.length).join(" ") : undefined);
1267
1554
  process.stdout.write("Approved " + taskId + " in " + runId + ". Resume explicitly with `forge resume " + runId + "`.\n");
1268
1555
  process.stdout.write("Run status: " + updated.status + "\n");
1269
1556
  }
@@ -1276,12 +1563,12 @@ Exit codes:
1276
1563
  case "resume": {
1277
1564
  const runId = rest[0] ?? "";
1278
1565
  const pathIndex = rest.indexOf("--path");
1279
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1566
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1280
1567
  try {
1281
1568
  if (!runId)
1282
1569
  throw new Error("Usage: anvil resume RUN_ID [--path PROJECT]");
1283
1570
  const permissions = await loadPermissionPolicy(target);
1284
- const providerChoice = (getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai")).toLowerCase();
1571
+ const providerChoice = getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai").toLowerCase();
1285
1572
  assertSupportedAgentPolicy(permissions, providerChoice, "ANVIL resume");
1286
1573
  await withRunExecutionLock(target, runId, async () => {
1287
1574
  const savedRun = await loadRun(target, runId);
@@ -1295,7 +1582,9 @@ Exit codes:
1295
1582
  throw new Error("Only pending or approval-cleared runs can be resumed. Use `anvil replay` only after reviewing failed tasks.");
1296
1583
  const model = getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ?? getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL");
1297
1584
  const baseModel = model ?? (providerChoice === "claude-code" ? "sonnet" : "gpt-4.1-mini");
1298
- const createProvider = (selectedModel) => providerChoice === "claude-code" ? new ClaudeCodeProvider({ cwd: target, model: selectedModel }) : createHostedAgentProvider({ ...process.env, ANVIL_AI_PROVIDER: providerChoice, FORGE_AI_PROVIDER: providerChoice }, selectedModel, { workspacePath: target, permissionPolicy: permissions });
1585
+ const createProvider = (selectedModel) => providerChoice === "claude-code"
1586
+ ? new ClaudeCodeProvider({ cwd: target, model: selectedModel })
1587
+ : createHostedAgentProvider({ ...process.env, ANVIL_AI_PROVIDER: providerChoice, FORGE_AI_PROVIDER: providerChoice }, selectedModel, { workspacePath: target, permissionPolicy: permissions });
1299
1588
  let priorTraceEntries = [];
1300
1589
  try {
1301
1590
  priorTraceEntries = (await loadExecutionTrace(target, runId)).entries;
@@ -1304,17 +1593,48 @@ Exit codes:
1304
1593
  if (!(error && typeof error === "object" && "code" in error && error.code === "ENOENT"))
1305
1594
  throw error;
1306
1595
  }
1307
- const traceAgentUsage = priorTraceEntries.reduce((usage, entry) => { usage[entry.agent] = (usage[entry.agent] ?? 0) + entry.inputTokensEstimate + entry.outputTokensEstimate; return usage; }, {});
1596
+ const traceAgentUsage = priorTraceEntries.reduce((usage, entry) => {
1597
+ usage[entry.agent] = (usage[entry.agent] ?? 0) + entry.inputTokensEstimate + entry.outputTokensEstimate;
1598
+ return usage;
1599
+ }, {});
1308
1600
  const traceRunUsage = priorTraceEntries.reduce((total, entry) => total + entry.inputTokensEstimate + entry.outputTokensEstimate, 0);
1309
1601
  const priorAgentUsage = savedRun.tokenUsage?.estimatedTokensByAgent ?? traceAgentUsage;
1310
1602
  const priorRunUsage = savedRun.tokenUsage?.estimatedTokensUsed ?? traceRunUsage;
1311
1603
  const resumeFallback = getAnvilEnv("ANVIL_FALLBACK_MODEL", "FORGE_FALLBACK_MODEL");
1312
1604
  const router = new ModelRouter({
1313
- fastModel: getAnvilEnv("ANVIL_MODEL_FAST", "FORGE_MODEL_FAST") ?? baseModel, balancedModel: getAnvilEnv("ANVIL_MODEL_BALANCED", "FORGE_MODEL_BALANCED") ?? baseModel, reasoningModel: getAnvilEnv("ANVIL_MODEL_REASONING", "FORGE_MODEL_REASONING") ?? baseModel, ...(resumeFallback ? { fallbackModel: resumeFallback } : {}), retryAttempts: Number(getAnvilEnv("ANVIL_MODEL_RETRY_ATTEMPTS", "FORGE_MODEL_RETRY_ATTEMPTS", "2")), retryDelayMs: Number(getAnvilEnv("ANVIL_MODEL_RETRY_DELAY_MS", "FORGE_MODEL_RETRY_DELAY_MS", "500")), ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK") ? { maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")) } : {}), ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK") ? { maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")) } : {}), ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN") ? { maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")) } : {}),
1314
- ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT") ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() } : {}),
1315
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS") ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) } : {}),
1316
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT") ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() } : {}),
1317
- initialEstimatedTokensUsed: priorRunUsage, initialEstimatedTokensByAgent: priorAgentUsage, createProvider
1605
+ fastModel: getAnvilEnv("ANVIL_MODEL_FAST", "FORGE_MODEL_FAST") ?? baseModel,
1606
+ balancedModel: getAnvilEnv("ANVIL_MODEL_BALANCED", "FORGE_MODEL_BALANCED") ?? baseModel,
1607
+ reasoningModel: getAnvilEnv("ANVIL_MODEL_REASONING", "FORGE_MODEL_REASONING") ?? baseModel,
1608
+ ...(resumeFallback ? { fallbackModel: resumeFallback } : {}),
1609
+ retryAttempts: Number(getAnvilEnv("ANVIL_MODEL_RETRY_ATTEMPTS", "FORGE_MODEL_RETRY_ATTEMPTS", "2")),
1610
+ retryDelayMs: Number(getAnvilEnv("ANVIL_MODEL_RETRY_DELAY_MS", "FORGE_MODEL_RETRY_DELAY_MS", "500")),
1611
+ ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")
1612
+ ? {
1613
+ maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")),
1614
+ }
1615
+ : {}),
1616
+ ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")
1617
+ ? {
1618
+ maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")),
1619
+ }
1620
+ : {}),
1621
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")
1622
+ ? {
1623
+ maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")),
1624
+ }
1625
+ : {}),
1626
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT")
1627
+ ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() }
1628
+ : {}),
1629
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")
1630
+ ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) }
1631
+ : {}),
1632
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT")
1633
+ ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() }
1634
+ : {}),
1635
+ initialEstimatedTokensUsed: priorRunUsage,
1636
+ initialEstimatedTokensByAgent: priorAgentUsage,
1637
+ createProvider,
1318
1638
  });
1319
1639
  const memory = await buildProjectMemoryContext(target, savedRun.goal);
1320
1640
  // Re-check current workspace policy before resuming any task that has not completed.
@@ -1324,10 +1644,12 @@ Exit codes:
1324
1644
  if (record.status !== "pending" && record.status !== "awaiting_approval")
1325
1645
  continue;
1326
1646
  const applied = applyBeforeAgentHooks(record.task, resumeHooks);
1327
- const additions = applied.context.filter(message => !record.task.prompt.includes(message));
1647
+ const additions = applied.context.filter((message) => !record.task.prompt.includes(message));
1328
1648
  record.task = {
1329
1649
  ...applied.task,
1330
- prompt: additions.length ? record.task.prompt + "\n\nHook guidance:\n" + additions.join("\n") : record.task.prompt,
1650
+ prompt: additions.length
1651
+ ? record.task.prompt + "\n\nHook guidance:\n" + additions.join("\n")
1652
+ : record.task.prompt,
1331
1653
  };
1332
1654
  if (applied.blockedReason) {
1333
1655
  record.status = "blocked";
@@ -1335,7 +1657,17 @@ Exit codes:
1335
1657
  }
1336
1658
  }
1337
1659
  const currentTokenUsage = () => {
1338
- const usage = router.traces.reduce((total, entry) => { total.estimatedTokensUsed += entry.inputTokensEstimate + entry.outputTokensEstimate; total.estimatedTokensByAgent[entry.agent] = (total.estimatedTokensByAgent[entry.agent] ?? 0) + entry.inputTokensEstimate + entry.outputTokensEstimate; return total; }, { estimatedTokensUsed: priorRunUsage, estimatedTokensByAgent: { ...priorAgentUsage } });
1660
+ const usage = router.traces.reduce((total, entry) => {
1661
+ total.estimatedTokensUsed += entry.inputTokensEstimate + entry.outputTokensEstimate;
1662
+ total.estimatedTokensByAgent[entry.agent] =
1663
+ (total.estimatedTokensByAgent[entry.agent] ?? 0) +
1664
+ entry.inputTokensEstimate +
1665
+ entry.outputTokensEstimate;
1666
+ return total;
1667
+ }, {
1668
+ estimatedTokensUsed: priorRunUsage,
1669
+ estimatedTokensByAgent: { ...priorAgentUsage },
1670
+ });
1339
1671
  return usage;
1340
1672
  };
1341
1673
  let persistedTraceEntries = 0;
@@ -1347,7 +1679,14 @@ Exit codes:
1347
1679
  persistedTraceEntries = router.traces.length;
1348
1680
  }
1349
1681
  };
1350
- const completed = await advanceRun(savedRun, router, new Set(), { maxConcurrency: 4, projectContext: memory.text, projectContextForTask: task => buildAgentWorkspaceContext(target, task), shouldCancel: async () => (await loadRun(target, runId)).status === "cancelled", onTaskSettled: persistRunSnapshot, onUpdate: persistRunSnapshot });
1682
+ const completed = await advanceRun(savedRun, router, new Set(), {
1683
+ maxConcurrency: 4,
1684
+ projectContext: memory.text,
1685
+ projectContextForTask: (task) => buildAgentWorkspaceContext(target, task),
1686
+ shouldCancel: async () => (await loadRun(target, runId)).status === "cancelled",
1687
+ onTaskSettled: persistRunSnapshot,
1688
+ onUpdate: persistRunSnapshot,
1689
+ });
1351
1690
  const saved = await saveRun(target, { ...completed, tokenUsage: currentTokenUsage() });
1352
1691
  const tracePath = await saveExecutionTrace(target, completed.id, router.traces.slice(persistedTraceEntries));
1353
1692
  process.stdout.write("Resumed " + runId + ": " + completed.status + "\n");
@@ -1363,7 +1702,7 @@ Exit codes:
1363
1702
  case "trace": {
1364
1703
  const runId = rest[0];
1365
1704
  const pathIndex = rest.indexOf("--path");
1366
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1705
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1367
1706
  if (!runId) {
1368
1707
  process.stderr.write("Usage: anvil trace RUNID [--path PROJECT]\n");
1369
1708
  process.exitCode = 1;
@@ -1373,11 +1712,49 @@ Exit codes:
1373
1712
  const trace = await loadExecutionTrace(target, runId);
1374
1713
  process.stdout.write("FORGE execution trace: " + runId + "\n");
1375
1714
  for (const entry of trace.entries)
1376
- process.stdout.write(" " + entry.taskId + " [" + entry.tier + "] " + entry.model + " | " + entry.durationMs + " ms | " + entry.outcome + " | ~" + entry.inputTokensEstimate + "/" + entry.outputTokensEstimate + " tokens\n");
1377
- process.stdout.write("Totals: " + trace.totals.tasks + " tasks, " + trace.totals.estimatedInputTokens + " input tokens, " + trace.totals.estimatedOutputTokens + " output tokens, " + trace.totals.fallbacks + " fallbacks, " + trace.totals.failures + " failures\n");
1715
+ process.stdout.write(" " +
1716
+ entry.taskId +
1717
+ " [" +
1718
+ entry.tier +
1719
+ "] " +
1720
+ entry.model +
1721
+ " | " +
1722
+ entry.durationMs +
1723
+ " ms | " +
1724
+ entry.outcome +
1725
+ " | ~" +
1726
+ entry.inputTokensEstimate +
1727
+ "/" +
1728
+ entry.outputTokensEstimate +
1729
+ " tokens\n");
1730
+ process.stdout.write("Totals: " +
1731
+ trace.totals.tasks +
1732
+ " tasks, " +
1733
+ trace.totals.estimatedInputTokens +
1734
+ " input tokens, " +
1735
+ trace.totals.estimatedOutputTokens +
1736
+ " output tokens, " +
1737
+ trace.totals.fallbacks +
1738
+ " fallbacks, " +
1739
+ trace.totals.failures +
1740
+ " failures\n");
1378
1741
  process.stdout.write("Provider health:\n");
1379
1742
  for (const provider of trace.providers ?? [])
1380
- process.stdout.write(" " + provider.name + ": " + provider.requests + " requests, " + provider.successes + " successes, " + provider.fallbacks + " fallbacks, " + provider.failures + " failures, " + provider.retries + " retries, " + provider.durationMs + " ms\n");
1743
+ process.stdout.write(" " +
1744
+ provider.name +
1745
+ ": " +
1746
+ provider.requests +
1747
+ " requests, " +
1748
+ provider.successes +
1749
+ " successes, " +
1750
+ provider.fallbacks +
1751
+ " fallbacks, " +
1752
+ provider.failures +
1753
+ " failures, " +
1754
+ provider.retries +
1755
+ " retries, " +
1756
+ provider.durationMs +
1757
+ " ms\n");
1381
1758
  }
1382
1759
  catch (error) {
1383
1760
  process.stderr.write("Could not read execution trace: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -1404,29 +1781,34 @@ Exit codes:
1404
1781
  throw new Error("Usage: anvil runs [path] [--json] [--limit N] [--status STATUS]");
1405
1782
  }
1406
1783
  for (const option of ["--json", "--limit", "--status"]) {
1407
- if (rest.filter(arg => arg === option).length > 1) {
1784
+ if (rest.filter((arg) => arg === option).length > 1) {
1408
1785
  throw new Error(`${option} may only be specified once.`);
1409
1786
  }
1410
1787
  }
1411
1788
  const allowedOptions = new Set(["--json", "--limit", "--status"]);
1412
- const unknownOption = rest.find(arg => arg.startsWith("--") && !allowedOptions.has(arg));
1789
+ const unknownOption = rest.find((arg) => arg.startsWith("--") && !allowedOptions.has(arg));
1413
1790
  if (unknownOption)
1414
1791
  throw new Error(`Unknown option: ${unknownOption}.`);
1415
1792
  if (limitIndex >= 0 && (!rawLimit || rawLimit.startsWith("--") || !Number.isSafeInteger(limit) || limit < 1)) {
1416
1793
  throw new Error("--limit requires a positive safe integer.");
1417
1794
  }
1418
1795
  const validStatuses = ["pending", "awaiting_approval", "running", "completed", "failed", "cancelled"];
1419
- if (statusIndex >= 0 && (!rawStatus || rawStatus.startsWith("--") || !validStatuses.includes(rawStatus))) {
1796
+ if (statusIndex >= 0 &&
1797
+ (!rawStatus ||
1798
+ rawStatus.startsWith("--") ||
1799
+ !validStatuses.includes(rawStatus))) {
1420
1800
  throw new Error(`--status requires one of: ${validStatuses.join(", ")}.`);
1421
1801
  }
1422
1802
  const allRuns = await listRuns(path.resolve(target));
1423
- const filteredRuns = rawStatus ? allRuns.filter(run => run.status === rawStatus) : allRuns;
1803
+ const filteredRuns = rawStatus ? allRuns.filter((run) => run.status === rawStatus) : allRuns;
1424
1804
  const runs = limit === undefined ? filteredRuns : filteredRuns.slice(0, limit);
1425
1805
  if (rest.includes("--json")) {
1426
1806
  process.stdout.write(formatRunHistoryJson(runs));
1427
1807
  }
1428
1808
  else {
1429
- process.stdout.write(runs.length ? runs.map(run => `${run.id} ${run.status.padEnd(18)} ${run.goal}`).join("\n") + "\n" : "No saved FORGE runs.\n");
1809
+ process.stdout.write(runs.length
1810
+ ? runs.map((run) => `${run.id} ${run.status.padEnd(18)} ${run.goal}`).join("\n") + "\n"
1811
+ : "No saved FORGE runs.\n");
1430
1812
  }
1431
1813
  }
1432
1814
  catch (error) {
@@ -1441,7 +1823,7 @@ Exit codes:
1441
1823
  process.stdout.write("Usage: anvil show-run RUN_ID [--path PROJECT]\nDisplays one validated saved run record.\n");
1442
1824
  return;
1443
1825
  }
1444
- const pathIndices = rest.flatMap((arg, index) => arg === "--path" ? [index] : []);
1826
+ const pathIndices = rest.flatMap((arg, index) => (arg === "--path" ? [index] : []));
1445
1827
  const pathIndex = pathIndices[0] ?? -1;
1446
1828
  const pathValue = pathIndex >= 0 ? rest[pathIndex + 1] : undefined;
1447
1829
  const positional = rest.filter((arg, index) => !arg.startsWith("--") && !(pathIndex >= 0 && index === pathIndex + 1));
@@ -1449,8 +1831,8 @@ Exit codes:
1449
1831
  try {
1450
1832
  if (pathIndices.length > 1)
1451
1833
  throw new Error("--path may only be specified once.");
1452
- if (rest.some(arg => arg.startsWith("--") && arg !== "--path")) {
1453
- throw new Error(`Unknown option: ${rest.find(arg => arg.startsWith("--") && arg !== "--path")}.`);
1834
+ if (rest.some((arg) => arg.startsWith("--") && arg !== "--path")) {
1835
+ throw new Error(`Unknown option: ${rest.find((arg) => arg.startsWith("--") && arg !== "--path")}.`);
1454
1836
  }
1455
1837
  if (pathIndex >= 0 && (!pathValue || pathValue.startsWith("--"))) {
1456
1838
  throw new Error("--path requires a project directory.");
@@ -1472,12 +1854,17 @@ Exit codes:
1472
1854
  const goalIndex = rest.indexOf("--goal");
1473
1855
  const pathIndex = rest.indexOf("--path");
1474
1856
  const goalEnd = pathIndex >= 0 ? pathIndex : rest.length;
1475
- const goal = goalIndex >= 0 ? rest.slice(goalIndex + 1, goalEnd).join(" ").trim() : "";
1476
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1857
+ const goal = goalIndex >= 0
1858
+ ? rest
1859
+ .slice(goalIndex + 1, goalEnd)
1860
+ .join(" ")
1861
+ .trim()
1862
+ : "";
1863
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1477
1864
  try {
1478
1865
  const permissions = await loadPermissionPolicy(target);
1479
1866
  const model = getAnvilEnv("ANVIL_AI_MODEL", "FORGE_AI_MODEL") ?? getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL");
1480
- const providerChoice = (getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai")).toLowerCase();
1867
+ const providerChoice = getAnvilEnv("ANVIL_AI_PROVIDER", "FORGE_AI_PROVIDER", "openai").toLowerCase();
1481
1868
  assertSupportedAgentPolicy(permissions, providerChoice, "ANVIL run");
1482
1869
  const baseModel = model ?? (providerChoice === "claude-code" ? "sonnet" : "gpt-4.1-mini");
1483
1870
  const apiKey = getAnvilEnv("ANVIL_API_KEY", "FORGE_API_KEY") ?? process.env.OPENAI_API_KEY ?? "";
@@ -1499,12 +1886,30 @@ Exit codes:
1499
1886
  retryAttempts: Number(getAnvilEnv("ANVIL_MODEL_RETRY_ATTEMPTS", "FORGE_MODEL_RETRY_ATTEMPTS", "2")),
1500
1887
  retryDelayMs: Number(getAnvilEnv("ANVIL_MODEL_RETRY_DELAY_MS", "FORGE_MODEL_RETRY_DELAY_MS", "500")),
1501
1888
  createProvider,
1502
- ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK") ? { maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")) } : {}),
1503
- ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK") ? { maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")) } : {}),
1504
- ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN") ? { maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")) } : {}),
1505
- ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT") ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() } : {}),
1506
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS") ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) } : {}),
1507
- ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT") ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() } : {}),
1889
+ ...(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")
1890
+ ? {
1891
+ maxInputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_INPUT_TOKENS_PER_TASK", "FORGE_MAX_INPUT_TOKENS_PER_TASK")),
1892
+ }
1893
+ : {}),
1894
+ ...(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")
1895
+ ? {
1896
+ maxOutputTokensPerTask: Number(getAnvilEnv("ANVIL_MAX_OUTPUT_TOKENS_PER_TASK", "FORGE_MAX_OUTPUT_TOKENS_PER_TASK")),
1897
+ }
1898
+ : {}),
1899
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")
1900
+ ? {
1901
+ maxEstimatedTokensPerRun: Number(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_RUN", "FORGE_MAX_ESTIMATED_TOKENS_PER_RUN")),
1902
+ }
1903
+ : {}),
1904
+ ...(getAnvilEnv("ANVIL_MAX_ESTIMATED_TOKENS_PER_AGENT", "FORGE_MAX_ESTIMATED_TOKENS_PER_AGENT")
1905
+ ? { maxEstimatedTokensPerAgent: readAgentTokenBudgets() }
1906
+ : {}),
1907
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")
1908
+ ? { maxTaskDurationMs: Number(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS", "FORGE_MAX_TASK_DURATION_MS")) }
1909
+ : {}),
1910
+ ...(getAnvilEnv("ANVIL_MAX_TASK_DURATION_MS_BY_AGENT", "FORGE_MAX_TASK_DURATION_MS_BY_AGENT")
1911
+ ? { maxTaskDurationMsByAgent: readAgentDurationBudgets() }
1912
+ : {}),
1508
1913
  });
1509
1914
  const provider = router;
1510
1915
  const plan = createReviewPlan(goal);
@@ -1513,7 +1918,12 @@ Exit codes:
1513
1918
  const hooks = await loadHooks(target);
1514
1919
  for (const record of plan.tasks) {
1515
1920
  const applied = applyBeforeAgentHooks(record.task, hooks);
1516
- record.task = { ...applied.task, prompt: applied.context.length ? applied.task.prompt + "\n\nHook guidance:\n" + applied.context.join("\n") : applied.task.prompt };
1921
+ record.task = {
1922
+ ...applied.task,
1923
+ prompt: applied.context.length
1924
+ ? applied.task.prompt + "\n\nHook guidance:\n" + applied.context.join("\n")
1925
+ : applied.task.prompt,
1926
+ };
1517
1927
  if (applied.blockedReason) {
1518
1928
  record.status = "blocked";
1519
1929
  record.error = "Blocked by policy hook: " + applied.blockedReason;
@@ -1521,16 +1931,25 @@ Exit codes:
1521
1931
  }
1522
1932
  process.stdout.write("Starting " + plan.id + " with provider " + provider.name + (model ? " (model " + model + ")" : "") + "\n");
1523
1933
  if (skills.length)
1524
- process.stdout.write("Selected skills: " + skills.map(skill => skill.id).join(", ") + "\n");
1934
+ process.stdout.write("Selected skills: " + skills.map((skill) => skill.id).join(", ") + "\n");
1525
1935
  if (hooks.length)
1526
1936
  process.stdout.write("Loaded policy hooks: " + hooks.length + "\n");
1527
1937
  const memory = await buildProjectMemoryContext(target, goal);
1528
1938
  const projectContext = [memory.text, skillContext].filter(Boolean).join("\n\n");
1529
- const completed = await advanceRun(plan, provider, new Set(), { maxConcurrency: 4, projectContext, projectContextForTask: task => buildAgentWorkspaceContext(target, task) });
1939
+ const completed = await advanceRun(plan, provider, new Set(), {
1940
+ maxConcurrency: 4,
1941
+ projectContext,
1942
+ projectContextForTask: (task) => buildAgentWorkspaceContext(target, task),
1943
+ });
1530
1944
  const saved = await saveRun(target, completed);
1531
1945
  const tracePath = await saveExecutionTrace(target, completed.id, router.traces);
1532
1946
  process.stdout.write(`Run status: ${completed.status}\nSaved history: ${path.relative(target, saved)}\nExecution trace: ${path.relative(target, tracePath)}\n`);
1533
- const totals = router.traces.reduce((sum, trace) => ({ input: sum.input + trace.inputTokensEstimate, output: sum.output + trace.outputTokensEstimate, duration: sum.duration + trace.durationMs, fallback: sum.fallback + (trace.outcome === "fallback" ? 1 : 0) }), { input: 0, output: 0, duration: 0, fallback: 0 });
1947
+ const totals = router.traces.reduce((sum, trace) => ({
1948
+ input: sum.input + trace.inputTokensEstimate,
1949
+ output: sum.output + trace.outputTokensEstimate,
1950
+ duration: sum.duration + trace.durationMs,
1951
+ fallback: sum.fallback + (trace.outcome === "fallback" ? 1 : 0),
1952
+ }), { input: 0, output: 0, duration: 0, fallback: 0 });
1534
1953
  process.stdout.write(`Estimated tokens: ${totals.input} in / ${totals.output} out | Agent time: ${totals.duration} ms | Fallbacks: ${totals.fallback}\n`);
1535
1954
  for (const task of completed.tasks)
1536
1955
  process.stdout.write(` ${task.status.padEnd(18)} ${task.task.agent}: ${task.task.title}\n`);
@@ -1549,11 +1968,17 @@ Exit codes:
1549
1968
  process.stdout.write("Usage: anvil deps [path] [--audit-report FILE] [--outdated-report FILE] [--json]\n");
1550
1969
  return;
1551
1970
  }
1552
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
1971
+ const option = (name) => {
1972
+ const index = rest.indexOf(name);
1973
+ return index >= 0 ? rest[index + 1] : undefined;
1974
+ };
1553
1975
  const target = path.resolve(rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--audit-report" && rest[index - 1] !== "--outdated-report") ?? ".");
1554
1976
  try {
1555
1977
  const auditReport = option("--audit-report"), outdatedReport = option("--outdated-report");
1556
- const report = await analyzeDependencies(target, { ...(auditReport ? { auditReport } : {}), ...(outdatedReport ? { outdatedReport } : {}) });
1978
+ const report = await analyzeDependencies(target, {
1979
+ ...(auditReport ? { auditReport } : {}),
1980
+ ...(outdatedReport ? { outdatedReport } : {}),
1981
+ });
1557
1982
  process.stdout.write(rest.includes("--json") ? JSON.stringify(report, null, 2) + "\n" : formatDependencyIntelligence(report) + "\n");
1558
1983
  }
1559
1984
  catch (error) {
@@ -1563,7 +1988,7 @@ Exit codes:
1563
1988
  return;
1564
1989
  }
1565
1990
  case "snapshot": {
1566
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
1991
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
1567
1992
  try {
1568
1993
  const comparison = await captureIntelligenceSnapshot(target);
1569
1994
  process.stdout.write(formatIntelligenceComparison(comparison) + "\n");
@@ -1575,10 +2000,12 @@ Exit codes:
1575
2000
  return;
1576
2001
  }
1577
2002
  case "timeline": {
1578
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
2003
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
1579
2004
  try {
1580
2005
  const comparison = await compareLatestIntelligence(target);
1581
- process.stdout.write(comparison ? formatIntelligenceComparison(comparison) + "\n" : "No engineering snapshots found. Run forge snapshot . first.\n");
2006
+ process.stdout.write(comparison
2007
+ ? formatIntelligenceComparison(comparison) + "\n"
2008
+ : "No engineering snapshots found. Run forge snapshot . first.\n");
1582
2009
  }
1583
2010
  catch (error) {
1584
2011
  process.stderr.write("FORGE timeline failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -1588,8 +2015,10 @@ Exit codes:
1588
2015
  }
1589
2016
  case "change-risk": {
1590
2017
  const pathIndex = rest.indexOf("--path"), baseIndex = rest.indexOf("--base");
1591
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? ".");
1592
- const base = baseIndex >= 0 ? rest[baseIndex + 1] ?? "HEAD" : "HEAD";
2018
+ const target = path.resolve(pathIndex >= 0
2019
+ ? (rest[pathIndex + 1] ?? ".")
2020
+ : (rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? "."));
2021
+ const base = baseIndex >= 0 ? (rest[baseIndex + 1] ?? "HEAD") : "HEAD";
1593
2022
  const json = rest.includes("--json");
1594
2023
  try {
1595
2024
  const report = await analyzeChangeIntelligence(target, base);
@@ -1603,7 +2032,7 @@ Exit codes:
1603
2032
  }
1604
2033
  case "risk": {
1605
2034
  const pathIndex = rest.indexOf("--path");
1606
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find(arg => !arg.startsWith("--")) ?? ".");
2035
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : (rest.find((arg) => !arg.startsWith("--")) ?? "."));
1607
2036
  const json = rest.includes("--json");
1608
2037
  try {
1609
2038
  const report = await analyzeEngineeringRisk(target);
@@ -1617,8 +2046,10 @@ Exit codes:
1617
2046
  }
1618
2047
  case "impact": {
1619
2048
  const pathIndex = rest.indexOf("--path"), baseIndex = rest.indexOf("--base");
1620
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? ".");
1621
- const base = baseIndex >= 0 ? rest[baseIndex + 1] ?? "HEAD" : "HEAD";
2049
+ const target = path.resolve(pathIndex >= 0
2050
+ ? (rest[pathIndex + 1] ?? ".")
2051
+ : (rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? "."));
2052
+ const base = baseIndex >= 0 ? (rest[baseIndex + 1] ?? "HEAD") : "HEAD";
1622
2053
  const json = rest.includes("--json");
1623
2054
  try {
1624
2055
  const report = await analyzeDependencyImpact(target, base);
@@ -1637,8 +2068,11 @@ Exit codes:
1637
2068
  optionValues.add(pathIndex + 1);
1638
2069
  if (limitIndex >= 0)
1639
2070
  optionValues.add(limitIndex + 1);
1640
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1641
- const query = rest.filter((arg, index) => !arg.startsWith("--") && !optionValues.has(index)).join(" ").trim();
2071
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2072
+ const query = rest
2073
+ .filter((arg, index) => !arg.startsWith("--") && !optionValues.has(index))
2074
+ .join(" ")
2075
+ .trim();
1642
2076
  const limit = limitIndex >= 0 ? Number(rest[limitIndex + 1]) : 80;
1643
2077
  try {
1644
2078
  const report = await searchWorkspace(target, query, limit);
@@ -1651,7 +2085,7 @@ Exit codes:
1651
2085
  return;
1652
2086
  }
1653
2087
  case "dashboard": {
1654
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
2088
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
1655
2089
  const portIndex = rest.indexOf("--port");
1656
2090
  const portValue = portIndex >= 0 ? Number(rest[portIndex + 1]) : 4173;
1657
2091
  if (!Number.isInteger(portValue) || portValue < 0 || portValue > 65535) {
@@ -1670,7 +2104,11 @@ Exit codes:
1670
2104
  break;
1671
2105
  }
1672
2106
  catch (err) {
1673
- if (err && typeof err === "object" && "code" in err && err.code === "EADDRINUSE" && p < portValue + 9)
2107
+ if (err &&
2108
+ typeof err === "object" &&
2109
+ "code" in err &&
2110
+ err.code === "EADDRINUSE" &&
2111
+ p < portValue + 9)
1674
2112
  continue;
1675
2113
  throw err;
1676
2114
  }
@@ -1698,12 +2136,14 @@ Exit codes:
1698
2136
  case "verify": {
1699
2137
  const listOnly = rest.includes("--list");
1700
2138
  const pathIndex = rest.indexOf("--path");
1701
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
1702
- const script = rest.find(arg => !arg.startsWith("--"));
2139
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2140
+ const script = rest.find((arg) => !arg.startsWith("--"));
1703
2141
  try {
1704
2142
  if (listOnly) {
1705
2143
  const checks = await discoverChecks(target);
1706
- process.stdout.write(checks.length ? checks.map(check => `${check.name.padEnd(20)} ${check.command}`).join("\n") + "\n" : "No selectable npm scripts found.\n");
2144
+ process.stdout.write(checks.length
2145
+ ? checks.map((check) => `${check.name.padEnd(20)} ${check.command}`).join("\n") + "\n"
2146
+ : "No selectable npm scripts found.\n");
1707
2147
  return;
1708
2148
  }
1709
2149
  if (!script)
@@ -1730,12 +2170,22 @@ Exit codes:
1730
2170
  for (const profile of profiles) {
1731
2171
  process.stdout.write([
1732
2172
  profile.name + " — " + profile.description,
1733
- " network=" + profile.network + ", readOnlyRoot=" + profile.readOnlyRoot +
1734
- ", dropAllCapabilities=" + profile.dropAllCapabilities +
1735
- ", noNewPrivileges=" + profile.noNewPrivileges,
1736
- " memory=" + profile.maxMemory + ", cpus=" + profile.maxCpus +
1737
- ", processes=" + profile.maxProcesses +
1738
- ", requireRootless=" + profile.requireRootless,
2173
+ " network=" +
2174
+ profile.network +
2175
+ ", readOnlyRoot=" +
2176
+ profile.readOnlyRoot +
2177
+ ", dropAllCapabilities=" +
2178
+ profile.dropAllCapabilities +
2179
+ ", noNewPrivileges=" +
2180
+ profile.noNewPrivileges,
2181
+ " memory=" +
2182
+ profile.maxMemory +
2183
+ ", cpus=" +
2184
+ profile.maxCpus +
2185
+ ", processes=" +
2186
+ profile.maxProcesses +
2187
+ ", requireRootless=" +
2188
+ profile.requireRootless,
1739
2189
  "",
1740
2190
  ].join("\n"));
1741
2191
  }
@@ -1759,13 +2209,17 @@ Exit codes:
1759
2209
  }
1760
2210
  case "sandbox-cleanup": {
1761
2211
  const apply = rest.includes("--apply");
1762
- const unknown = rest.filter(arg => arg !== "--apply");
2212
+ const unknown = rest.filter((arg) => arg !== "--apply");
1763
2213
  try {
1764
2214
  if (unknown.length)
1765
2215
  throw new Error("Usage: anvil sandbox-cleanup [--apply]. Cleanup is dry-run unless --apply is explicit.");
1766
2216
  const output = await runSandboxProbe([
1767
- "ps", "-a", "--filter", "label=com.forge.managed=true",
1768
- "--format", "{{.ID}}\\t{{.Status}}\\t{{.Names}}",
2217
+ "ps",
2218
+ "-a",
2219
+ "--filter",
2220
+ "label=com.forge.managed=true",
2221
+ "--format",
2222
+ "{{.ID}}\\t{{.Status}}\\t{{.Names}}",
1769
2223
  ]);
1770
2224
  const managed = parseManagedContainers(output);
1771
2225
  const candidates = selectStoppedManagedContainers(managed);
@@ -1858,7 +2312,7 @@ Exit codes:
1858
2312
  }
1859
2313
  case "patch-verify": {
1860
2314
  const pathIndex = rest.indexOf("--path");
1861
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2315
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1862
2316
  const installDependencies = rest.includes("--install-dependencies");
1863
2317
  const sandboxImageIndex = rest.indexOf("--sandbox-image");
1864
2318
  const sandboxImage = sandboxImageIndex >= 0 ? rest[sandboxImageIndex + 1] : undefined;
@@ -1867,8 +2321,17 @@ Exit codes:
1867
2321
  const securityProfile = profileIndex >= 0 ? rest[profileIndex + 1] : undefined;
1868
2322
  const policyIndex = rest.indexOf("--policy");
1869
2323
  const sandboxPolicyPath = policyIndex >= 0 ? rest[policyIndex + 1] : undefined;
1870
- const argsWithoutPath = rest.filter((arg, index) => arg !== "--path" && !(pathIndex >= 0 && index === pathIndex + 1) && arg !== "--install-dependencies" && arg !== "--sandbox-image" && !(sandboxImageIndex >= 0 && index === sandboxImageIndex + 1) && arg !== "--require-rootless" && arg !== "--profile" && !(profileIndex >= 0 && index === profileIndex + 1) && arg !== "--policy" && !(policyIndex >= 0 && index === policyIndex + 1));
1871
- const stageId = argsWithoutPath.find(arg => !arg.startsWith("--"));
2324
+ const argsWithoutPath = rest.filter((arg, index) => arg !== "--path" &&
2325
+ !(pathIndex >= 0 && index === pathIndex + 1) &&
2326
+ arg !== "--install-dependencies" &&
2327
+ arg !== "--sandbox-image" &&
2328
+ !(sandboxImageIndex >= 0 && index === sandboxImageIndex + 1) &&
2329
+ arg !== "--require-rootless" &&
2330
+ arg !== "--profile" &&
2331
+ !(profileIndex >= 0 && index === profileIndex + 1) &&
2332
+ arg !== "--policy" &&
2333
+ !(policyIndex >= 0 && index === policyIndex + 1));
2334
+ const stageId = argsWithoutPath.find((arg) => !arg.startsWith("--"));
1872
2335
  const script = argsWithoutPath.find((arg, index) => !arg.startsWith("--") && index > 0);
1873
2336
  try {
1874
2337
  if (!stageId || !script)
@@ -1891,21 +2354,51 @@ Exit codes:
1891
2354
  installDependencies,
1892
2355
  ...(sandboxImage ? { sandboxImage } : {}),
1893
2356
  ...(requireRootless ? { requireRootless } : {}),
1894
- ...(securityProfile ? { securityProfile: securityProfile } : {}),
2357
+ ...(securityProfile
2358
+ ? { securityProfile: securityProfile }
2359
+ : {}),
1895
2360
  ...(sandboxPolicyPath ? { sandboxPolicyPath } : {}),
1896
2361
  });
1897
2362
  process.stdout.write("# ANVIL Staged Patch Verification\n\n" +
1898
- "- **Stage:** `" + result.stageId + "`\n" +
1899
- "- **Patch file:** `" + result.file + "`\n" +
1900
- "- **Staged SHA-256:** `" + result.stagedSha256 + "`\n" +
1901
- "- **Snapshot files:** `" + result.snapshotFileCount + "`\n" +
1902
- "- **Snapshot bytes:** `" + result.snapshotBytes + "`\n" +
1903
- (result.dependencyInstall ? "- **Dependencies:** **" + result.dependencyInstall.status.toUpperCase() + "** (`" + result.dependencyInstall.command + "`)\n" : "") +
1904
- (sandboxImage ? "- **Execution:** **CONTAINER** (`" + sandboxImage + "`)\n" : "- **Execution:** **LOCAL HOST**\n") +
1905
- "- **Verification:** **" + result.verification.status.toUpperCase() + "**\n" +
1906
- "- **Duration:** `" + result.verification.durationMs + " ms`\n\n" +
1907
- "## Output\n\n```\n" + (result.verification.output || "(No output captured.)") + "\n```\n\n" +
1908
- "Temporary workspace was removed after verification. " + (sandboxImage ? "The container runtime and host kernel remain trusted." : "Project scripts ran on the local host; this is not an OS-level sandbox.") + "\n");
2363
+ "- **Stage:** `" +
2364
+ result.stageId +
2365
+ "`\n" +
2366
+ "- **Patch file:** `" +
2367
+ result.file +
2368
+ "`\n" +
2369
+ "- **Staged SHA-256:** `" +
2370
+ result.stagedSha256 +
2371
+ "`\n" +
2372
+ "- **Snapshot files:** `" +
2373
+ result.snapshotFileCount +
2374
+ "`\n" +
2375
+ "- **Snapshot bytes:** `" +
2376
+ result.snapshotBytes +
2377
+ "`\n" +
2378
+ (result.dependencyInstall
2379
+ ? "- **Dependencies:** **" +
2380
+ result.dependencyInstall.status.toUpperCase() +
2381
+ "** (`" +
2382
+ result.dependencyInstall.command +
2383
+ "`)\n"
2384
+ : "") +
2385
+ (sandboxImage
2386
+ ? "- **Execution:** **CONTAINER** (`" + sandboxImage + "`)\n"
2387
+ : "- **Execution:** **LOCAL HOST**\n") +
2388
+ "- **Verification:** **" +
2389
+ result.verification.status.toUpperCase() +
2390
+ "**\n" +
2391
+ "- **Duration:** `" +
2392
+ result.verification.durationMs +
2393
+ " ms`\n\n" +
2394
+ "## Output\n\n```\n" +
2395
+ (result.verification.output || "(No output captured.)") +
2396
+ "\n```\n\n" +
2397
+ "Temporary workspace was removed after verification. " +
2398
+ (sandboxImage
2399
+ ? "The container runtime and host kernel remain trusted."
2400
+ : "Project scripts ran on the local host; this is not an OS-level sandbox.") +
2401
+ "\n");
1909
2402
  if (result.verification.status !== "passed")
1910
2403
  process.exitCode = 1;
1911
2404
  }
@@ -1918,14 +2411,16 @@ Exit codes:
1918
2411
  }
1919
2412
  case "verify-history": {
1920
2413
  const targetIndex = rest.indexOf("--path");
1921
- const target = path.resolve(targetIndex >= 0 ? rest[targetIndex + 1] ?? "." : ".");
2414
+ const target = path.resolve(targetIndex >= 0 ? (rest[targetIndex + 1] ?? ".") : ".");
1922
2415
  try {
1923
2416
  const history = await listVerifications(target);
1924
2417
  if (!history.length) {
1925
2418
  process.stdout.write("No saved verification runs.\n");
1926
2419
  return;
1927
2420
  }
1928
- process.stdout.write(history.map(item => `${item.startedAt} ${item.status.padEnd(12)} ${item.script.padEnd(16)} ${item.durationMs} ms ${item.id}`).join("\n") + "\n");
2421
+ process.stdout.write(history
2422
+ .map((item) => `${item.startedAt} ${item.status.padEnd(12)} ${item.script.padEnd(16)} ${item.durationMs} ms ${item.id}`)
2423
+ .join("\n") + "\n");
1929
2424
  }
1930
2425
  catch (error) {
1931
2426
  const message = error instanceof Error ? error.message : String(error);
@@ -1939,8 +2434,10 @@ Exit codes:
1939
2434
  const all = rest.includes("--all");
1940
2435
  const pathIndex = rest.indexOf("--path");
1941
2436
  const baseIndex = rest.indexOf("--base");
1942
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base" && rest[index - 1] !== "--path") ?? ".");
1943
- const base = baseIndex >= 0 ? rest[baseIndex + 1] ?? "HEAD" : "HEAD";
2437
+ const target = path.resolve(pathIndex >= 0
2438
+ ? (rest[pathIndex + 1] ?? ".")
2439
+ : (rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base" && rest[index - 1] !== "--path") ?? "."));
2440
+ const base = baseIndex >= 0 ? (rest[baseIndex + 1] ?? "HEAD") : "HEAD";
1944
2441
  try {
1945
2442
  const report = await reviewProject(target, { base, all });
1946
2443
  let ai;
@@ -1948,11 +2445,21 @@ Exit codes:
1948
2445
  const apiKey = getAnvilEnv("ANVIL_API_KEY", "FORGE_API_KEY") ?? process.env.OPENAI_API_KEY ?? "";
1949
2446
  if (!apiKey)
1950
2447
  throw new Error("Set ANVIL_API_KEY, FORGE_API_KEY, or OPENAI_API_KEY before using --ai.");
1951
- const provider = new OpenAICompatibleProvider({ apiKey, model: getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL", "gpt-4.1-mini"), baseUrl: getAnvilEnv("ANVIL_BASE_URL", "FORGE_BASE_URL", "https://api.openai.com/v1") });
2448
+ const provider = new OpenAICompatibleProvider({
2449
+ apiKey,
2450
+ model: getAnvilEnv("ANVIL_MODEL", "FORGE_MODEL", "gpt-4.1-mini"),
2451
+ baseUrl: getAnvilEnv("ANVIL_BASE_URL", "FORGE_BASE_URL", "https://api.openai.com/v1"),
2452
+ });
1952
2453
  ai = await generateAIReview(report, provider);
1953
2454
  }
1954
2455
  const saved = await saveReview(target, report, ai);
1955
- process.stdout.write(json ? JSON.stringify({ ...saved, reviewId: saved.id }, null, 2) + "\n" : formatCodeReview(report) + (ai ? "\n\n" + formatAIReview(ai) : "") + "\n\nSaved review: " + saved.id + "\n");
2456
+ process.stdout.write(json
2457
+ ? JSON.stringify({ ...saved, reviewId: saved.id }, null, 2) + "\n"
2458
+ : formatCodeReview(report) +
2459
+ (ai ? "\n\n" + formatAIReview(ai) : "") +
2460
+ "\n\nSaved review: " +
2461
+ saved.id +
2462
+ "\n");
1956
2463
  }
1957
2464
  catch (error) {
1958
2465
  const message = error instanceof Error ? error.message : String(error);
@@ -1965,7 +2472,11 @@ Exit codes:
1965
2472
  const target = path.resolve(rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path") ?? ".");
1966
2473
  try {
1967
2474
  const records = await listReviews(target);
1968
- process.stdout.write(records.length ? records.map(r => r.id + " " + r.createdAt + " " + r.report.findings.length + " static findings").join("\n") + "\n" : "No saved reviews found.\n");
2475
+ process.stdout.write(records.length
2476
+ ? records
2477
+ .map((r) => r.id + " " + r.createdAt + " " + r.report.findings.length + " static findings")
2478
+ .join("\n") + "\n"
2479
+ : "No saved reviews found.\n");
1969
2480
  }
1970
2481
  catch (error) {
1971
2482
  process.stderr.write("ANVIL review history failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -1974,9 +2485,9 @@ Exit codes:
1974
2485
  return;
1975
2486
  }
1976
2487
  case "review-show": {
1977
- const id = rest.find(arg => !arg.startsWith("--")) ?? "";
2488
+ const id = rest.find((arg) => !arg.startsWith("--")) ?? "";
1978
2489
  const pathIndex = rest.indexOf("--path");
1979
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2490
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1980
2491
  try {
1981
2492
  const record = await loadReview(target, id);
1982
2493
  process.stdout.write(formatReviewRecord(record) + "\n");
@@ -1990,13 +2501,21 @@ Exit codes:
1990
2501
  case "review-compare": {
1991
2502
  const ids = rest.filter((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path");
1992
2503
  const pathIndex = rest.indexOf("--path");
1993
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2504
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
1994
2505
  try {
1995
2506
  if (ids.length < 2)
1996
2507
  throw new Error("Usage: anvil review-compare REVIEW_ID_A REVIEW_ID_B --path PROJECT");
1997
2508
  const [a, b] = await Promise.all([loadReview(target, ids[0]), loadReview(target, ids[1])]);
1998
2509
  const diff = compareReviews(a, b);
1999
- process.stdout.write(["Review comparison", "Added (" + diff.added.length + ")", ...diff.added.map(x => "+ " + x), "Resolved (" + diff.resolved.length + ")", ...diff.resolved.map(x => "- " + x), "Persistent (" + diff.persistent.length + ")", ...diff.persistent.map(x => "= " + x)].join("\n") + "\n");
2510
+ process.stdout.write([
2511
+ "Review comparison",
2512
+ "Added (" + diff.added.length + ")",
2513
+ ...diff.added.map((x) => "+ " + x),
2514
+ "Resolved (" + diff.resolved.length + ")",
2515
+ ...diff.resolved.map((x) => "- " + x),
2516
+ "Persistent (" + diff.persistent.length + ")",
2517
+ ...diff.persistent.map((x) => "= " + x),
2518
+ ].join("\n") + "\n");
2000
2519
  }
2001
2520
  catch (error) {
2002
2521
  process.stderr.write("ANVIL review comparison failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -2007,8 +2526,10 @@ Exit codes:
2007
2526
  case "regression": {
2008
2527
  const pathIndex = rest.indexOf("--path");
2009
2528
  const baseIndex = rest.indexOf("--base");
2010
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? ".");
2011
- const base = baseIndex >= 0 ? rest[baseIndex + 1] ?? "HEAD" : "HEAD";
2529
+ const target = path.resolve(pathIndex >= 0
2530
+ ? (rest[pathIndex + 1] ?? ".")
2531
+ : (rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--base") ?? "."));
2532
+ const base = baseIndex >= 0 ? (rest[baseIndex + 1] ?? "HEAD") : "HEAD";
2012
2533
  const json = rest.includes("--json");
2013
2534
  try {
2014
2535
  const report = await generateRegressionReport(target, base);
@@ -2024,13 +2545,22 @@ Exit codes:
2024
2545
  }
2025
2546
  case "patch-review": {
2026
2547
  const pathIndex = rest.indexOf("--path");
2027
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2548
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2028
2549
  const stageId = rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path") ?? "";
2029
2550
  try {
2030
2551
  if (!stageId)
2031
2552
  throw new Error("Usage: anvil patch-review STAGE_ID --path PROJECT");
2032
2553
  const review = await reviewStagedPatch(target, stageId);
2033
- process.stdout.write(["# ANVIL Staged Patch Review", "- Stage: " + review.stageId, "- File: " + review.file, "- Source SHA-256: " + review.sourceSha256, "- Staged SHA-256: " + review.stagedSha256, "- Source unchanged: " + review.unchanged, "", review.diff].join("\n") + "\n");
2554
+ process.stdout.write([
2555
+ "# ANVIL Staged Patch Review",
2556
+ "- Stage: " + review.stageId,
2557
+ "- File: " + review.file,
2558
+ "- Source SHA-256: " + review.sourceSha256,
2559
+ "- Staged SHA-256: " + review.stagedSha256,
2560
+ "- Source unchanged: " + review.unchanged,
2561
+ "",
2562
+ review.diff,
2563
+ ].join("\n") + "\n");
2034
2564
  }
2035
2565
  catch (error) {
2036
2566
  process.stderr.write("ANVIL staged patch review failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -2040,13 +2570,23 @@ Exit codes:
2040
2570
  }
2041
2571
  case "patch-stage": {
2042
2572
  const pathIndex = rest.indexOf("--path");
2043
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2573
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2044
2574
  const approvalId = rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path") ?? "";
2045
2575
  try {
2046
2576
  if (!approvalId)
2047
2577
  throw new Error("Usage: anvil patch-stage APPROVAL_ID --path PROJECT");
2048
2578
  const staged = await stageApprovedPatch(target, approvalId);
2049
- process.stdout.write(["# ANVIL Isolated Patch Stage", "- Stage: " + staged.id, "- Approval: " + staged.approvalId, "- File: " + staged.file, "- Staged copy: " + staged.stagedFile, "- SHA-256: " + staged.stagedSha256, "", "The original project file was not modified.", "No tests or commands were executed."].join("\n") + "\n");
2579
+ process.stdout.write([
2580
+ "# ANVIL Isolated Patch Stage",
2581
+ "- Stage: " + staged.id,
2582
+ "- Approval: " + staged.approvalId,
2583
+ "- File: " + staged.file,
2584
+ "- Staged copy: " + staged.stagedFile,
2585
+ "- SHA-256: " + staged.stagedSha256,
2586
+ "",
2587
+ "The original project file was not modified.",
2588
+ "No tests or commands were executed.",
2589
+ ].join("\n") + "\n");
2050
2590
  }
2051
2591
  catch (error) {
2052
2592
  process.stderr.write("ANVIL patch staging failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -2058,18 +2598,31 @@ Exit codes:
2058
2598
  const pathIndex = rest.indexOf("--path");
2059
2599
  const reviewerIndex = rest.indexOf("--reviewer");
2060
2600
  const reasonIndex = rest.indexOf("--reason");
2061
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2601
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2062
2602
  const values = rest.filter((arg, index) => !arg.startsWith("--") && index !== pathIndex + 1 && index !== reviewerIndex + 1 && index !== reasonIndex + 1);
2063
2603
  const reviewId = values[0] ?? "";
2064
2604
  const suggestionNumber = Number(values[1]);
2065
- const reviewer = reviewerIndex >= 0 ? rest[reviewerIndex + 1] ?? "" : "";
2605
+ const reviewer = reviewerIndex >= 0 ? (rest[reviewerIndex + 1] ?? "") : "";
2066
2606
  const reason = reasonIndex >= 0 ? rest[reasonIndex + 1] : undefined;
2067
2607
  try {
2068
2608
  if (!reviewId || !Number.isInteger(suggestionNumber) || suggestionNumber < 1 || !reviewer.trim()) {
2069
2609
  throw new Error("Usage: anvil patch-approve REVIEW_ID SUGGESTION_NUMBER --reviewer NAME [--reason TEXT] --path PROJECT");
2070
2610
  }
2071
2611
  const approval = await approveReviewPatch(target, reviewId, suggestionNumber, reviewer, reason);
2072
- process.stdout.write(["# ANVIL Patch Approval", "- Approval ID: " + approval.id, "- Review: " + approval.reviewId, "- Suggestion: " + approval.suggestionNumber, "- File: " + approval.file, "- Reviewer: " + approval.reviewer, "- Approved: " + approval.approvedAt, "- Expires: " + approval.expiresAt, "- Patch SHA-256: " + approval.patchSha256, "- Source SHA-256: " + approval.sourceSha256, "", "No source files were modified."].join("\n") + "\n");
2612
+ process.stdout.write([
2613
+ "# ANVIL Patch Approval",
2614
+ "- Approval ID: " + approval.id,
2615
+ "- Review: " + approval.reviewId,
2616
+ "- Suggestion: " + approval.suggestionNumber,
2617
+ "- File: " + approval.file,
2618
+ "- Reviewer: " + approval.reviewer,
2619
+ "- Approved: " + approval.approvedAt,
2620
+ "- Expires: " + approval.expiresAt,
2621
+ "- Patch SHA-256: " + approval.patchSha256,
2622
+ "- Source SHA-256: " + approval.sourceSha256,
2623
+ "",
2624
+ "No source files were modified.",
2625
+ ].join("\n") + "\n");
2073
2626
  }
2074
2627
  catch (error) {
2075
2628
  process.stderr.write("ANVIL patch approval failed: " + (error instanceof Error ? error.message : String(error)) + "\n");
@@ -2080,7 +2633,7 @@ Exit codes:
2080
2633
  case "patch-approval-check": {
2081
2634
  const id = rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path") ?? "";
2082
2635
  const pathIndex = rest.indexOf("--path");
2083
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2636
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2084
2637
  try {
2085
2638
  if (!id)
2086
2639
  throw new Error("Usage: anvil patch-approval-check APPROVAL_ID --path PROJECT");
@@ -2098,7 +2651,7 @@ Exit codes:
2098
2651
  case "patch-check": {
2099
2652
  const id = rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--path") ?? "";
2100
2653
  const pathIndex = rest.indexOf("--path");
2101
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : ".");
2654
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : ".");
2102
2655
  const numberArg = rest.find((arg, index) => index > 0 && !arg.startsWith("--") && rest[index - 1] !== id && rest[index - 1] !== "--path");
2103
2656
  try {
2104
2657
  if (!id || !numberArg)
@@ -2119,8 +2672,8 @@ Exit codes:
2119
2672
  const json = rest.includes("--json");
2120
2673
  const pathIndex = rest.indexOf("--path");
2121
2674
  const baseIndex = rest.indexOf("--base");
2122
- const target = path.resolve(pathIndex >= 0 ? rest[pathIndex + 1] ?? "." : rest.find(arg => !arg.startsWith("--")) ?? ".");
2123
- const base = baseIndex >= 0 ? rest[baseIndex + 1] ?? "HEAD" : "HEAD";
2675
+ const target = path.resolve(pathIndex >= 0 ? (rest[pathIndex + 1] ?? ".") : (rest.find((arg) => !arg.startsWith("--")) ?? "."));
2676
+ const base = baseIndex >= 0 ? (rest[baseIndex + 1] ?? "HEAD") : "HEAD";
2124
2677
  try {
2125
2678
  const report = await analyzeGitChanges(target, base);
2126
2679
  process.stdout.write(json ? JSON.stringify(report, null, 2) + "\n" : formatGitIntelligence(report) + "\n");
@@ -2134,7 +2687,7 @@ Exit codes:
2134
2687
  }
2135
2688
  case "deep": {
2136
2689
  const json = rest.includes("--json");
2137
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
2690
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
2138
2691
  try {
2139
2692
  const report = await analyzeCodebase(target);
2140
2693
  process.stdout.write(json ? JSON.stringify(report, null, 2) + "\n" : formatCodebaseAnalysis(report) + "\n");
@@ -2151,11 +2704,17 @@ Exit codes:
2151
2704
  process.stdout.write("Usage: anvil remediate [path] [--audit-report FILE] [--outdated-report FILE] [--json]\n");
2152
2705
  return;
2153
2706
  }
2154
- const option = (name) => { const index = rest.indexOf(name); return index >= 0 ? rest[index + 1] : undefined; };
2707
+ const option = (name) => {
2708
+ const index = rest.indexOf(name);
2709
+ return index >= 0 ? rest[index + 1] : undefined;
2710
+ };
2155
2711
  const target = path.resolve(rest.find((arg, index) => !arg.startsWith("--") && rest[index - 1] !== "--audit-report" && rest[index - 1] !== "--outdated-report") ?? ".");
2156
2712
  try {
2157
2713
  const auditReport = option("--audit-report"), outdatedReport = option("--outdated-report");
2158
- const plan = await generateRemediationPlan(target, { ...(auditReport ? { auditReport } : {}), ...(outdatedReport ? { outdatedReport } : {}) });
2714
+ const plan = await generateRemediationPlan(target, {
2715
+ ...(auditReport ? { auditReport } : {}),
2716
+ ...(outdatedReport ? { outdatedReport } : {}),
2717
+ });
2159
2718
  process.stdout.write(rest.includes("--json") ? JSON.stringify(plan, null, 2) + "\n" : formatRemediationPlan(plan) + "\n");
2160
2719
  }
2161
2720
  catch (error) {
@@ -2166,7 +2725,7 @@ Exit codes:
2166
2725
  }
2167
2726
  case "health": {
2168
2727
  const json = rest.includes("--json");
2169
- const target = path.resolve(rest.find(arg => !arg.startsWith("--")) ?? ".");
2728
+ const target = path.resolve(rest.find((arg) => !arg.startsWith("--")) ?? ".");
2170
2729
  try {
2171
2730
  const report = await generateProjectHealth(target);
2172
2731
  process.stdout.write(json ? JSON.stringify(report, null, 2) + "\n" : formatProjectHealthMarkdown(report) + "\n");
@@ -2190,7 +2749,7 @@ Exit codes:
2190
2749
  const fix = isFix || rest.includes("--fix");
2191
2750
  const verify = rest.includes("--verify");
2192
2751
  const checkOrphans = !rest.includes("--no-orphans");
2193
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
2752
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
2194
2753
  try {
2195
2754
  const { cleanProject, formatCleanReport } = await import("./core/project-cleaner.js");
2196
2755
  const report = await cleanProject(path.resolve(target), { apply, fix, checkOrphans, verify });
@@ -2209,7 +2768,7 @@ Exit codes:
2209
2768
  return;
2210
2769
  }
2211
2770
  const json = rest.includes("--json");
2212
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
2771
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
2213
2772
  try {
2214
2773
  const { undoLastClean, formatUndoReport } = await import("./core/project-cleaner.js");
2215
2774
  const report = await undoLastClean(path.resolve(target));
@@ -2231,7 +2790,7 @@ Exit codes:
2231
2790
  return;
2232
2791
  }
2233
2792
  const json = rest.includes("--json");
2234
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
2793
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
2235
2794
  try {
2236
2795
  const report = await auditProject(path.resolve(target));
2237
2796
  process.stdout.write(json ? `${JSON.stringify(report, null, 2)}\n` : `${formatQualityAudit(report)}\n`);
@@ -2250,7 +2809,7 @@ Exit codes:
2250
2809
  return;
2251
2810
  }
2252
2811
  const json = rest.includes("--json");
2253
- const target = rest.find(arg => !arg.startsWith("--")) ?? ".";
2812
+ const target = rest.find((arg) => !arg.startsWith("--")) ?? ".";
2254
2813
  try {
2255
2814
  if (command === "inspect") {
2256
2815
  const report = await inspectProject(path.resolve(target));
@@ -2277,7 +2836,7 @@ Exit codes:
2277
2836
  }
2278
2837
  const json = rest.includes("--json");
2279
2838
  const doFix = rest.includes("--fix");
2280
- const target = rest.find(arg => !arg.startsWith("--"));
2839
+ const target = rest.find((arg) => !arg.startsWith("--"));
2281
2840
  try {
2282
2841
  const { runDoctorChecks, fixDoctorIssues, formatDoctorReport } = await import("./core/doctor.js");
2283
2842
  if (doFix) {