ai-engineering-loop 2.2.0 → 2.3.0
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.
- package/.agents/skills/ai-engineering-loop/SKILL.md +4 -4
- package/.agents/skills/ai-engineering-loop/references/loop-deliver.md +2 -2
- package/.agents/skills/ai-engineering-loop/references/loop-goal.md +3 -3
- package/.agents/skills/ai-engineering-loop/references/loop-maker.md +1 -0
- package/.agents/skills/ai-engineering-loop/references/loop-review.md +3 -4
- package/.agents/skills/ai-engineering-loop/references/loop-verify.md +3 -3
- package/.claude/agents/judge.md +5 -2
- package/.grok/agents/judge.md +5 -2
- package/CHANGELOG.md +35 -0
- package/agents/maker.md +1 -1
- package/agents/shared/judge.body.md +5 -2
- package/bin/ai-engineering-loop.js +303 -77
- package/core/definition-of-done.md +2 -2
- package/core/escalation-policy.md +1 -1
- package/core/goal-contract.md +1 -1
- package/core/iteration-policy.md +2 -1
- package/core/judge-policy.md +10 -4
- package/core/orchestration-model.md +8 -7
- package/core/repo-config-schema.md +2 -2
- package/core/runtime-safety.md +3 -3
- package/lib/auto-update.js +8 -14
- package/lib/claimed-vs-reality.js +2 -2
- package/lib/cli-args.js +3 -1
- package/lib/docs.js +1 -1
- package/lib/doctor.js +0 -6
- package/lib/escalation.js +1 -1
- package/lib/fsutil.js +88 -0
- package/lib/gates.js +69 -27
- package/lib/generate-workflow.js +0 -29
- package/lib/goal-runtime.js +3 -2
- package/lib/guard-hooks.js +26 -7
- package/lib/guard.js +100 -10
- package/lib/hosts/doctor.js +1 -5
- package/lib/impact-map.js +4 -9
- package/lib/jsonl-ledger.js +104 -0
- package/lib/learn.js +4 -9
- package/lib/loop-groups.js +0 -25
- package/lib/migrations.js +1 -5
- package/lib/no-progress.js +119 -0
- package/lib/node-evidence.js +1 -5
- package/lib/orchestration.js +5 -361
- package/lib/personal-workflow.js +1 -1
- package/lib/quick.js +152 -0
- package/lib/receipt.js +1 -5
- package/lib/recipe-builder.js +3 -35
- package/lib/recipe-graph.js +2 -63
- package/lib/recipe.js +2 -4
- package/lib/review-ingest.js +116 -0
- package/lib/run-abandon.js +2 -1
- package/lib/run-compaction.js +4 -12
- package/lib/run-interactions.js +16 -71
- package/lib/run-lifecycle.js +25 -77
- package/lib/run-state.js +68 -17
- package/lib/run-watcher.js +11 -14
- package/lib/runtime-files.js +45 -1
- package/lib/safe-context.js +17 -6
- package/lib/sandbox.js +2 -1
- package/lib/smart-context.js +3 -13
- package/lib/spawn-command.js +94 -0
- package/lib/studio-layout.js +2 -8
- package/lib/studio-runtime.js +2 -1
- package/lib/studio-server.js +25 -5
- package/lib/tier.js +36 -10
- package/lib/user-state.js +2 -17
- package/lib/verification-lanes.js +115 -0
- package/lib/verification-recorder.js +110 -22
- package/lib/workflow-runtime.js +7 -4
- package/package.json +1 -2
- package/policies/review-budget.md +1 -1
- package/.claude/skills/ai-engineering-loop/SKILL.md +0 -62
- package/.claude/skills/ai-engineering-loop/references/loop-commands.md +0 -24
- package/.claude/skills/ai-engineering-loop/references/loop-deliver.md +0 -10
- package/.claude/skills/ai-engineering-loop/references/loop-goal.md +0 -11
- package/.claude/skills/ai-engineering-loop/references/loop-hosts.md +0 -48
- package/.claude/skills/ai-engineering-loop/references/loop-maker.md +0 -14
- package/.claude/skills/ai-engineering-loop/references/loop-review.md +0 -20
- package/.claude/skills/ai-engineering-loop/references/loop-risky.md +0 -9
- package/.claude/skills/ai-engineering-loop/references/loop-verify.md +0 -20
- package/.claude/skills/animate/RECIPES.md +0 -324
- package/.claude/skills/animate/SKILL.md +0 -207
- package/.claude/skills/animate-expo/RECIPES.md +0 -385
- package/.claude/skills/animate-expo/SKILL.md +0 -263
- package/.claude/skills/animation-vocabulary/SKILL.md +0 -181
- package/.claude/skills/apple-design/SKILL.md +0 -290
- package/.claude/skills/ask-sonner/API.md +0 -64
- package/.claude/skills/ask-sonner/SKILL.md +0 -88
- package/.claude/skills/backend-development/SKILL.md +0 -272
- package/.claude/skills/backend-development/references/api-design-guidelines.md +0 -95
- package/.claude/skills/backend-development/references/query-and-database-best-practices.md +0 -53
- package/.claude/skills/backend-development/references/security-checklist.md +0 -60
- package/.claude/skills/emil-design-eng/SKILL.md +0 -674
- package/.claude/skills/find-animation-opportunities/SKILL.md +0 -140
- package/.claude/skills/improve-animations/AUDIT.md +0 -115
- package/.claude/skills/improve-animations/PLAN-TEMPLATE.md +0 -73
- package/.claude/skills/improve-animations/SKILL.md +0 -109
- package/.claude/skills/mobile-native/SKILL.md +0 -311
- package/.claude/skills/pick-ui-library/SKILL.md +0 -85
- package/.claude/skills/prototype/PICKER.md +0 -197
- package/.claude/skills/prototype/SKILL.md +0 -98
- package/.claude/skills/review-animations/SKILL.md +0 -120
- package/.claude/skills/review-animations/STANDARDS.md +0 -187
- package/.claude/skills/write-swift/SKILL.md +0 -396
- package/lib/intern-catalog.js +0 -53
package/lib/orchestration.js
CHANGED
|
@@ -17,7 +17,6 @@
|
|
|
17
17
|
* and must NEVER be used as the Devil's Advocate or Judge.
|
|
18
18
|
*/
|
|
19
19
|
|
|
20
|
-
const crypto = require('crypto');
|
|
21
20
|
const fs = require('fs');
|
|
22
21
|
const path = require('path');
|
|
23
22
|
|
|
@@ -66,219 +65,12 @@ const EXECUTION_MODE_ALIASES = {
|
|
|
66
65
|
ARTIFACT_ISOLATED_REVIEW: 'CONTEXT_ISOLATION_ONLY'
|
|
67
66
|
};
|
|
68
67
|
|
|
69
|
-
const GROK_FORBIDDEN_REVIEW_TYPES = [
|
|
70
|
-
'caveman:cavecrew-reviewer',
|
|
71
|
-
'caveman:cavecrew-builder',
|
|
72
|
-
'caveman:cavecrew-investigator',
|
|
73
|
-
'explore',
|
|
74
|
-
'plan'
|
|
75
|
-
];
|
|
76
|
-
|
|
77
68
|
function resolveExecutionModeId(modeId) {
|
|
78
69
|
if (!modeId) return 'UNAVAILABLE';
|
|
79
70
|
return EXECUTION_MODE_ALIASES[modeId] || modeId;
|
|
80
71
|
}
|
|
81
|
-
|
|
82
|
-
/**
|
|
83
|
-
* 2. Standardized Capability Evidence Factory with 3-Stage Lifecycle
|
|
84
|
-
*/
|
|
85
|
-
function createCapabilityEvidence({
|
|
86
|
-
mechanism,
|
|
87
|
-
classification,
|
|
88
|
-
configurationSupported = false,
|
|
89
|
-
invocationAvailable = false,
|
|
90
|
-
executionProven = false,
|
|
91
|
-
available = false,
|
|
92
|
-
commandOrApi = null,
|
|
93
|
-
result = null,
|
|
94
|
-
executionIdentity = null,
|
|
95
|
-
conversationId = null,
|
|
96
|
-
parentConversationId = null,
|
|
97
|
-
childConversationId = null,
|
|
98
|
-
childModelResponse = null,
|
|
99
|
-
modelExecutionProven = false,
|
|
100
|
-
independentContextProven = false,
|
|
101
|
-
historyInherited = null,
|
|
102
|
-
isDocumentationOnly = false,
|
|
103
|
-
isBrowserAutomationOnly = false,
|
|
104
|
-
reason = null,
|
|
105
|
-
timestamp = new Date().toISOString()
|
|
106
|
-
}) {
|
|
107
|
-
if (!mechanism || !classification) {
|
|
108
|
-
throw new Error('Capability evidence must include mechanism and classification');
|
|
109
|
-
}
|
|
110
|
-
|
|
111
|
-
const modelExecuted = Boolean(modelExecutionProven || (childModelResponse && executionIdentity));
|
|
112
|
-
const fullyProven = Boolean(executionProven || (modelExecuted && childConversationId && independentContextProven));
|
|
113
|
-
|
|
114
|
-
return {
|
|
115
|
-
mechanism,
|
|
116
|
-
classification,
|
|
117
|
-
configurationSupported: Boolean(configurationSupported),
|
|
118
|
-
invocationAvailable: Boolean(invocationAvailable),
|
|
119
|
-
executionProven: fullyProven,
|
|
120
|
-
available: Boolean(available || fullyProven),
|
|
121
|
-
commandOrApi,
|
|
122
|
-
result,
|
|
123
|
-
executionIdentity,
|
|
124
|
-
conversationId,
|
|
125
|
-
parentConversationId,
|
|
126
|
-
childConversationId,
|
|
127
|
-
childModelResponse,
|
|
128
|
-
modelExecutionProven: modelExecuted,
|
|
129
|
-
independentContextProven: Boolean(independentContextProven),
|
|
130
|
-
historyInherited,
|
|
131
|
-
isDocumentationOnly: Boolean(isDocumentationOnly),
|
|
132
|
-
isBrowserAutomationOnly: Boolean(isBrowserAutomationOnly),
|
|
133
|
-
reason,
|
|
134
|
-
timestamp
|
|
135
|
-
};
|
|
136
|
-
}
|
|
137
|
-
|
|
138
72
|
/**
|
|
139
|
-
*
|
|
140
|
-
*/
|
|
141
|
-
function isCapabilityProvenForMode(evidence, targetModeId) {
|
|
142
|
-
if (!evidence) {
|
|
143
|
-
return false;
|
|
144
|
-
}
|
|
145
|
-
|
|
146
|
-
// Reject non-agent capabilities (such as IPC send-message)
|
|
147
|
-
if (
|
|
148
|
-
evidence.classification === 'NOT_AN_AGENT_EXECUTION_CAPABILITY' ||
|
|
149
|
-
evidence.classification === 'IPC_MESSAGE_DISPATCH'
|
|
150
|
-
) {
|
|
151
|
-
return false;
|
|
152
|
-
}
|
|
153
|
-
|
|
154
|
-
// Reject browser automation tools (browser_subagent is NOT an LLM subagent)
|
|
155
|
-
if (evidence.isBrowserAutomationOnly || evidence.classification === 'BROWSER_AUTOMATION_TOOL') {
|
|
156
|
-
return false;
|
|
157
|
-
}
|
|
158
|
-
|
|
159
|
-
// Reject compressed review presets that do not emit the Finding Ledger schema
|
|
160
|
-
if (
|
|
161
|
-
evidence.classification === 'GROK_COMPRESSED_REVIEW_PRESET' ||
|
|
162
|
-
GROK_FORBIDDEN_REVIEW_TYPES.includes(evidence.mechanism) ||
|
|
163
|
-
GROK_FORBIDDEN_REVIEW_TYPES.includes(evidence.commandOrApi)
|
|
164
|
-
) {
|
|
165
|
-
return false;
|
|
166
|
-
}
|
|
167
|
-
|
|
168
|
-
// Reject documentation-only or configuration-only claims
|
|
169
|
-
if (evidence.isDocumentationOnly || (evidence.configurationSupported && !evidence.executionProven)) {
|
|
170
|
-
if (targetModeId !== 'CONTEXT_ISOLATION_ONLY') {
|
|
171
|
-
return false;
|
|
172
|
-
}
|
|
173
|
-
}
|
|
174
|
-
|
|
175
|
-
// Modes requiring true independent LLM execution
|
|
176
|
-
if (
|
|
177
|
-
targetModeId === 'TRUE_INDEPENDENT_AGENT' ||
|
|
178
|
-
targetModeId === 'ISOLATED_AGENT_INSTANCE' ||
|
|
179
|
-
targetModeId === 'FRESH_PROCESS_AGENT'
|
|
180
|
-
) {
|
|
181
|
-
// Must satisfy all 4 pillars of execution proof:
|
|
182
|
-
const hasChildSession = Boolean(evidence.childConversationId || targetModeId === 'FRESH_PROCESS_AGENT');
|
|
183
|
-
const hasModelResponse = Boolean(evidence.modelExecutionProven || evidence.childModelResponse);
|
|
184
|
-
const hasExecIdentity = Boolean(evidence.executionIdentity);
|
|
185
|
-
const hasCleanContext = Boolean(evidence.independentContextProven && evidence.historyInherited !== true);
|
|
186
|
-
|
|
187
|
-
if (!evidence.executionProven && !(hasChildSession && hasModelResponse && hasExecIdentity && hasCleanContext)) {
|
|
188
|
-
return false;
|
|
189
|
-
}
|
|
190
|
-
}
|
|
191
|
-
|
|
192
|
-
// Context isolation mode requires verified artifact boundary
|
|
193
|
-
if (targetModeId === 'CONTEXT_ISOLATION_ONLY') {
|
|
194
|
-
return evidence.available === true;
|
|
195
|
-
}
|
|
196
|
-
|
|
197
|
-
return true;
|
|
198
|
-
}
|
|
199
|
-
|
|
200
|
-
/**
|
|
201
|
-
* 4. Deterministic Execution Mode Selection
|
|
202
|
-
* Evaluates registry strictly in priority order:
|
|
203
|
-
* 1. TRUE_INDEPENDENT_AGENT
|
|
204
|
-
* 2. ISOLATED_AGENT_INSTANCE
|
|
205
|
-
* 3. FRESH_PROCESS_AGENT
|
|
206
|
-
* 4. CONTEXT_ISOLATION_ONLY
|
|
207
|
-
* 5. UNAVAILABLE
|
|
208
|
-
*/
|
|
209
|
-
function selectExecutionMode(capabilityRegistry = {}) {
|
|
210
|
-
const {
|
|
211
|
-
nativeSubagent,
|
|
212
|
-
sdkAgent,
|
|
213
|
-
headlessProcessAgent,
|
|
214
|
-
artifactIsolation
|
|
215
|
-
} = capabilityRegistry;
|
|
216
|
-
|
|
217
|
-
// Priority 1: True Native Independent Sub-Agent
|
|
218
|
-
if (isCapabilityProvenForMode(nativeSubagent, 'TRUE_INDEPENDENT_AGENT')) {
|
|
219
|
-
return EXECUTION_MODES.TRUE_INDEPENDENT_AGENT;
|
|
220
|
-
}
|
|
221
|
-
|
|
222
|
-
// Priority 2: Programmatic SDK Agent Instance
|
|
223
|
-
if (isCapabilityProvenForMode(sdkAgent, 'ISOLATED_AGENT_INSTANCE')) {
|
|
224
|
-
return EXECUTION_MODES.ISOLATED_AGENT_INSTANCE;
|
|
225
|
-
}
|
|
226
|
-
|
|
227
|
-
// Priority 3: Headless Subprocess Agent
|
|
228
|
-
if (isCapabilityProvenForMode(headlessProcessAgent, 'FRESH_PROCESS_AGENT')) {
|
|
229
|
-
return EXECUTION_MODES.FRESH_PROCESS_AGENT;
|
|
230
|
-
}
|
|
231
|
-
|
|
232
|
-
// Priority 4: Guaranteed Fallback - Context Isolation Only
|
|
233
|
-
if (isCapabilityProvenForMode(artifactIsolation, 'CONTEXT_ISOLATION_ONLY')) {
|
|
234
|
-
return EXECUTION_MODES.CONTEXT_ISOLATION_ONLY;
|
|
235
|
-
}
|
|
236
|
-
|
|
237
|
-
// Priority 5: Unavailable
|
|
238
|
-
return EXECUTION_MODES.UNAVAILABLE;
|
|
239
|
-
}
|
|
240
|
-
|
|
241
|
-
/**
|
|
242
|
-
* 5. Build Clean-Slate Artifact Isolation Barrier
|
|
243
|
-
*/
|
|
244
|
-
function buildReviewContextBarrier({
|
|
245
|
-
goalContract,
|
|
246
|
-
gitDiff,
|
|
247
|
-
verificationLogs,
|
|
248
|
-
projectContext,
|
|
249
|
-
activeIteration = 1,
|
|
250
|
-
priorFindingSignatures = []
|
|
251
|
-
}) {
|
|
252
|
-
if (!goalContract || !gitDiff) {
|
|
253
|
-
throw new Error('Goal contract and git diff are mandatory for review barrier');
|
|
254
|
-
}
|
|
255
|
-
|
|
256
|
-
return {
|
|
257
|
-
iteration: activeIteration,
|
|
258
|
-
diffHash: crypto.createHash('sha256').update(gitDiff).digest('hex').slice(0, 16),
|
|
259
|
-
goalContract: {
|
|
260
|
-
objective: goalContract.objective,
|
|
261
|
-
acceptanceCriteria: goalContract.acceptanceCriteria || [],
|
|
262
|
-
technicalConstraints: goalContract.technicalConstraints || [],
|
|
263
|
-
outOfScope: goalContract.outOfScope || []
|
|
264
|
-
},
|
|
265
|
-
projectContext: {
|
|
266
|
-
profile: projectContext?.profile || 'standard',
|
|
267
|
-
architecture: projectContext?.architecture || '',
|
|
268
|
-
conventions: projectContext?.conventions || '',
|
|
269
|
-
verificationCommands: projectContext?.verificationCommands || ''
|
|
270
|
-
},
|
|
271
|
-
gitDiff: gitDiff.trim(),
|
|
272
|
-
verificationLogs: {
|
|
273
|
-
exitCode: verificationLogs?.exitCode ?? 0,
|
|
274
|
-
summary: verificationLogs?.summary || 'Deterministic checks passed 100%'
|
|
275
|
-
},
|
|
276
|
-
priorFindingSignatures: [...priorFindingSignatures]
|
|
277
|
-
};
|
|
278
|
-
}
|
|
279
|
-
|
|
280
|
-
/**
|
|
281
|
-
* 6. Validate Verification Evidence Contract
|
|
73
|
+
* 2. Validate Verification Evidence Contract
|
|
282
74
|
*/
|
|
283
75
|
function validateVerificationEvidence(evidence) {
|
|
284
76
|
if (!evidence) return { valid: false, reason: 'Evidence object missing' };
|
|
@@ -310,7 +102,7 @@ function validateVerificationEvidence(evidence) {
|
|
|
310
102
|
}
|
|
311
103
|
|
|
312
104
|
/**
|
|
313
|
-
*
|
|
105
|
+
* 3. Validate Finding Ledger Schema
|
|
314
106
|
*/
|
|
315
107
|
function validateFindingLedger(ledger) {
|
|
316
108
|
if (!ledger || !Array.isArray(ledger.findings)) {
|
|
@@ -365,7 +157,7 @@ function isBlockingFinding(finding) {
|
|
|
365
157
|
}
|
|
366
158
|
|
|
367
159
|
/**
|
|
368
|
-
*
|
|
160
|
+
* 4. Compute Judge Verdict based primarily on Validity + Severity, then review axis
|
|
369
161
|
*/
|
|
370
162
|
function computeJudgeVerdict({
|
|
371
163
|
goalContract,
|
|
@@ -432,7 +224,7 @@ function computeJudgeVerdict({
|
|
|
432
224
|
}
|
|
433
225
|
|
|
434
226
|
/**
|
|
435
|
-
*
|
|
227
|
+
* 5. Detect Grok CLI as a host runtime.
|
|
436
228
|
* Presence of the grok binary is CONFIGURATION_SUPPORTED.
|
|
437
229
|
* GROK_SUBAGENTS=0 means invocation is disabled even if grok is installed.
|
|
438
230
|
* Execution is never proven from detection alone.
|
|
@@ -484,161 +276,13 @@ function detectGrokRuntime(env = process.env, fsApi = fs) {
|
|
|
484
276
|
};
|
|
485
277
|
}
|
|
486
278
|
|
|
487
|
-
/**
|
|
488
|
-
* 10. Build capability evidence for Grok spawn_subagent.
|
|
489
|
-
* Independent context is proven only when the child did not resume a Maker transcript.
|
|
490
|
-
*/
|
|
491
|
-
function createGrokCapabilityEvidence({
|
|
492
|
-
invocationAvailable = false,
|
|
493
|
-
executionProven = false,
|
|
494
|
-
childConversationId = null,
|
|
495
|
-
childModelResponse = null,
|
|
496
|
-
executionIdentity = null,
|
|
497
|
-
independentContextProven = false,
|
|
498
|
-
historyInherited = null,
|
|
499
|
-
resumeFrom = null,
|
|
500
|
-
reason = null
|
|
501
|
-
} = {}) {
|
|
502
|
-
const resumedMaker = Boolean(resumeFrom);
|
|
503
|
-
const inherited = historyInherited === true || resumedMaker;
|
|
504
|
-
const modelExecuted = Boolean(executionProven || childModelResponse);
|
|
505
|
-
const fullyProven = Boolean(
|
|
506
|
-
modelExecuted &&
|
|
507
|
-
childConversationId &&
|
|
508
|
-
executionIdentity &&
|
|
509
|
-
independentContextProven &&
|
|
510
|
-
!inherited
|
|
511
|
-
);
|
|
512
|
-
|
|
513
|
-
return createCapabilityEvidence({
|
|
514
|
-
mechanism: 'grok-spawn_subagent',
|
|
515
|
-
classification: fullyProven
|
|
516
|
-
? 'TRUE_INDEPENDENT_AGENT'
|
|
517
|
-
: invocationAvailable
|
|
518
|
-
? 'INVOCATION_AVAILABLE'
|
|
519
|
-
: 'CONFIGURATION_SUPPORTED_WITHOUT_INVOCATION_TOOL',
|
|
520
|
-
configurationSupported: true,
|
|
521
|
-
invocationAvailable: Boolean(invocationAvailable),
|
|
522
|
-
executionProven: fullyProven,
|
|
523
|
-
available: Boolean(invocationAvailable || fullyProven),
|
|
524
|
-
commandOrApi: 'spawn_subagent',
|
|
525
|
-
childConversationId,
|
|
526
|
-
childModelResponse,
|
|
527
|
-
executionIdentity,
|
|
528
|
-
modelExecutionProven: modelExecuted,
|
|
529
|
-
independentContextProven: Boolean(independentContextProven && !inherited),
|
|
530
|
-
historyInherited: inherited ? true : historyInherited,
|
|
531
|
-
reason: reason || (resumedMaker
|
|
532
|
-
? 'resume_from inherits the source transcript; Devil\'s Advocate and Judge must spawn fresh'
|
|
533
|
-
: null)
|
|
534
|
-
});
|
|
535
|
-
}
|
|
536
|
-
|
|
537
|
-
/**
|
|
538
|
-
* 11. Grok spawn plan for Devil's Advocate / Judge.
|
|
539
|
-
* Parent orchestrator spawns children; children must not spawn children (depth 1).
|
|
540
|
-
* Never pass resume_from. Never use cavecrew-reviewer (wrong output schema).
|
|
541
|
-
*/
|
|
542
|
-
function buildGrokReviewSpawnPlan({
|
|
543
|
-
role,
|
|
544
|
-
iteration = 1,
|
|
545
|
-
artifactPaths = {}
|
|
546
|
-
} = {}) {
|
|
547
|
-
if (role !== 'devil-advocate' && role !== 'judge') {
|
|
548
|
-
throw new Error('Grok review spawn role must be devil-advocate or judge');
|
|
549
|
-
}
|
|
550
|
-
|
|
551
|
-
return {
|
|
552
|
-
tool: 'spawn_subagent',
|
|
553
|
-
subagent_type: role,
|
|
554
|
-
fallback_subagent_type: 'general-purpose',
|
|
555
|
-
description: role === 'judge'
|
|
556
|
-
? `[judge] evaluate iteration ${iteration}`
|
|
557
|
-
: `[devil-advocate] review iteration ${iteration}`,
|
|
558
|
-
background: false,
|
|
559
|
-
capability_mode: 'execute',
|
|
560
|
-
isolation: 'none',
|
|
561
|
-
resume_from: null,
|
|
562
|
-
forbidden_types: [...GROK_FORBIDDEN_REVIEW_TYPES],
|
|
563
|
-
artifactPaths: {
|
|
564
|
-
goalContract: artifactPaths.goalContract || null,
|
|
565
|
-
gitDiff: artifactPaths.gitDiff || null,
|
|
566
|
-
verificationLogs: artifactPaths.verificationLogs || null,
|
|
567
|
-
projectContext: artifactPaths.projectContext || '.ai-engineering-loop/',
|
|
568
|
-
findingLedger: artifactPaths.findingLedger || null
|
|
569
|
-
}
|
|
570
|
-
};
|
|
571
|
-
}
|
|
572
|
-
|
|
573
|
-
/**
|
|
574
|
-
* 12. Truthful Review Report Generator
|
|
575
|
-
* Enforces mandatory reporting headers and strictly prevents misleading phrasing.
|
|
576
|
-
* When a non-independent mode is reported although native subagent invocation
|
|
577
|
-
* is AVAILABLE, the Execution Mode line must carry the reason (#57), so the
|
|
578
|
-
* header never looks contradictory.
|
|
579
|
-
*/
|
|
580
|
-
const DEFAULT_ISOLATION_REASON = 'subagent available but no reviewer subagent returned its own result';
|
|
581
|
-
|
|
582
|
-
function formatExecutionReport({ selectedMode, capabilityRegistry = {}, reason = null }) {
|
|
583
|
-
const isContextOnly = selectedMode.id === EXECUTION_MODES.CONTEXT_ISOLATION_ONLY.id;
|
|
584
|
-
const nativeAvailable = Boolean(capabilityRegistry.nativeSubagent?.invocationAvailable);
|
|
585
|
-
const nativeSubagentState = nativeAvailable ? 'AVAILABLE' : 'UNAVAILABLE';
|
|
586
|
-
const needsReason = !selectedMode.isIndependentExecutionProven && (nativeAvailable || reason);
|
|
587
|
-
const why = String(reason || '').trim() || DEFAULT_ISOLATION_REASON;
|
|
588
|
-
|
|
589
|
-
const lines = [
|
|
590
|
-
`Execution Mode: ${selectedMode.id}${needsReason ? ` (reason: ${why})` : ''}`,
|
|
591
|
-
`Independent LLM Execution: ${selectedMode.isIndependentExecutionProven ? 'PROVEN' : 'NOT PROVEN'}`,
|
|
592
|
-
`Native Subagent Invocation: ${nativeSubagentState}`,
|
|
593
|
-
`Review Method: ${isContextOnly ? 'Clean-Slate Artifact Isolation Barrier' : selectedMode.label}`
|
|
594
|
-
];
|
|
595
|
-
|
|
596
|
-
return lines.join('\n');
|
|
597
|
-
}
|
|
598
|
-
|
|
599
|
-
/**
|
|
600
|
-
* 13. REPORT_ONLY (recipe `audit`) review policy (#57).
|
|
601
|
-
* A report-only audit has no Maker diff and no Judge, but its findings still
|
|
602
|
-
* benefit from one independent challenge when it is cheap: if the host can
|
|
603
|
-
* invoke a native subagent, spawn ONE fresh `devil-advocate` sibling on
|
|
604
|
-
* report.json before `gate report` (no Judge, no iteration loop). Otherwise the
|
|
605
|
-
* same check runs in-session as CONTEXT_ISOLATION_ONLY with an explicit reason.
|
|
606
|
-
*/
|
|
607
|
-
function reportOnlyReviewPlan(capabilityRegistry = {}) {
|
|
608
|
-
const nativeAvailable = Boolean(capabilityRegistry.nativeSubagent?.invocationAvailable);
|
|
609
|
-
if (nativeAvailable) {
|
|
610
|
-
return {
|
|
611
|
-
spawn: 'devil-advocate',
|
|
612
|
-
judge: false,
|
|
613
|
-
expectedMode: EXECUTION_MODES.TRUE_INDEPENDENT_AGENT.id,
|
|
614
|
-
reason: null
|
|
615
|
-
};
|
|
616
|
-
}
|
|
617
|
-
return {
|
|
618
|
-
spawn: null,
|
|
619
|
-
judge: false,
|
|
620
|
-
expectedMode: EXECUTION_MODES.CONTEXT_ISOLATION_ONLY.id,
|
|
621
|
-
reason: 'no native subagent tool in this session'
|
|
622
|
-
};
|
|
623
|
-
}
|
|
624
|
-
|
|
625
279
|
module.exports = {
|
|
626
280
|
EXECUTION_MODES,
|
|
627
281
|
EXECUTION_MODE_ALIASES,
|
|
628
|
-
GROK_FORBIDDEN_REVIEW_TYPES,
|
|
629
282
|
resolveExecutionModeId,
|
|
630
|
-
createCapabilityEvidence,
|
|
631
|
-
isCapabilityProvenForMode,
|
|
632
|
-
selectExecutionMode,
|
|
633
|
-
buildReviewContextBarrier,
|
|
634
283
|
validateVerificationEvidence,
|
|
635
284
|
validateFindingLedger,
|
|
636
285
|
isBlockingFinding,
|
|
637
286
|
computeJudgeVerdict,
|
|
638
|
-
|
|
639
|
-
reportOnlyReviewPlan,
|
|
640
|
-
DEFAULT_ISOLATION_REASON,
|
|
641
|
-
detectGrokRuntime,
|
|
642
|
-
createGrokCapabilityEvidence,
|
|
643
|
-
buildGrokReviewSpawnPlan
|
|
287
|
+
detectGrokRuntime
|
|
644
288
|
};
|
package/lib/personal-workflow.js
CHANGED
|
@@ -6,6 +6,7 @@
|
|
|
6
6
|
|
|
7
7
|
const fs = require('fs');
|
|
8
8
|
const path = require('path');
|
|
9
|
+
const { ACTOR_ID } = require('./fsutil.js');
|
|
9
10
|
const { redactHomePaths, redactSecrets } = require('./safe-context.js');
|
|
10
11
|
const {
|
|
11
12
|
homeDir,
|
|
@@ -41,7 +42,6 @@ const SUGGEST_RATIO = 0.8;
|
|
|
41
42
|
const RISK_WINDOW = 10;
|
|
42
43
|
const RISK_COUNT = 4;
|
|
43
44
|
const LEARNING_MIN = 3;
|
|
44
|
-
const ACTOR_ID = /^[A-Za-z0-9@._-]{1,64}$/;
|
|
45
45
|
const LEARNING_ID = /^[a-z0-9][a-z0-9-]{0,63}$/;
|
|
46
46
|
const GATE_ID = /^[a-z0-9-]{1,40}$/;
|
|
47
47
|
const BACKBONE_TYPES = new Set([
|
package/lib/quick.js
ADDED
|
@@ -0,0 +1,152 @@
|
|
|
1
|
+
'use strict';
|
|
2
|
+
|
|
3
|
+
// `quick start|finish`: the T0 path in two calls. Each step is the regular
|
|
4
|
+
// CLI command run as a child process, so every gate behaves exactly as in the
|
|
5
|
+
// long form; quick only removes the hand-written glue between them.
|
|
6
|
+
|
|
7
|
+
const fs = require('fs');
|
|
8
|
+
const path = require('path');
|
|
9
|
+
const { spawnSync } = require('child_process');
|
|
10
|
+
const { extractChangedPaths } = require('./escalation.js');
|
|
11
|
+
|
|
12
|
+
const CLI = path.join(__dirname, '..', 'bin', 'ai-engineering-loop.js');
|
|
13
|
+
const CLAIMED_PATH = '.ai-engineering-loop/tasks/claimed-vs-reality.md';
|
|
14
|
+
|
|
15
|
+
function quickError(message, code, details = []) {
|
|
16
|
+
return Object.assign(new Error(message), { code, details });
|
|
17
|
+
}
|
|
18
|
+
|
|
19
|
+
function invoke(rootDir, args) {
|
|
20
|
+
const result = spawnSync(process.execPath, [CLI, ...args, '--json'], {
|
|
21
|
+
cwd: rootDir,
|
|
22
|
+
env: { ...process.env, AEL_NO_AUTO_UPDATE: '1' },
|
|
23
|
+
encoding: 'utf8',
|
|
24
|
+
maxBuffer: 64 * 1024 * 1024
|
|
25
|
+
});
|
|
26
|
+
const lines = String(result.stdout || '').trim().split('\n');
|
|
27
|
+
let parsed = null;
|
|
28
|
+
for (let index = lines.length - 1; index >= 0 && !parsed; index -= 1) {
|
|
29
|
+
try {
|
|
30
|
+
parsed = JSON.parse(lines.slice(index).join('\n'));
|
|
31
|
+
} catch {
|
|
32
|
+
// keep widening until the trailing JSON document parses
|
|
33
|
+
}
|
|
34
|
+
}
|
|
35
|
+
if (result.status !== 0 || !parsed || parsed.ok === false) {
|
|
36
|
+
const message = parsed?.error || String(result.stderr || result.stdout || 'command failed').trim().split('\n').pop();
|
|
37
|
+
throw quickError(`\`${args.slice(0, 2).join(' ')}\` failed: ${message}`, parsed?.code || 'QUICK_STEP_FAILED', parsed?.details || []);
|
|
38
|
+
}
|
|
39
|
+
return parsed;
|
|
40
|
+
}
|
|
41
|
+
|
|
42
|
+
function oneLine(value, limit = 300) {
|
|
43
|
+
return String(value || '').replace(/\|/g, '/').replace(/\s+/g, ' ').trim().slice(0, limit);
|
|
44
|
+
}
|
|
45
|
+
|
|
46
|
+
function quickGoal(runId, task) {
|
|
47
|
+
return {
|
|
48
|
+
schemaVersion: 1,
|
|
49
|
+
runId,
|
|
50
|
+
objective: oneLine(task, 500),
|
|
51
|
+
acceptanceCriteria: [{
|
|
52
|
+
id: 'AC-1',
|
|
53
|
+
statement: `The change does exactly what the task says: ${oneLine(task, 200)}`,
|
|
54
|
+
evidenceRequired: 'verify command output (exit 0) and the gated diff',
|
|
55
|
+
failureCases: [
|
|
56
|
+
'files unrelated to the task are changed',
|
|
57
|
+
'the verification command fails'
|
|
58
|
+
]
|
|
59
|
+
}],
|
|
60
|
+
derivedBy: 'quick start (T0)'
|
|
61
|
+
};
|
|
62
|
+
}
|
|
63
|
+
|
|
64
|
+
function startQuick(rootDir, { task, mode = null } = {}) {
|
|
65
|
+
if (!String(task || '').trim()) throw quickError('quick start needs a task: quick start "<task>"', 'INVALID_USAGE');
|
|
66
|
+
const run = invoke(rootDir, ['run', '--new', ...(mode ? ['--mode', mode] : []), task]);
|
|
67
|
+
if (run.tier !== 'T0') {
|
|
68
|
+
return {
|
|
69
|
+
ok: true,
|
|
70
|
+
runId: run.runId,
|
|
71
|
+
tier: run.tier,
|
|
72
|
+
quick: false,
|
|
73
|
+
learnings: run.learnings || [],
|
|
74
|
+
next: `Tier ${run.tier} is not trivial; continue the normal loop for Run ${run.runId} (Goal Contract, Maker, verify, review).`
|
|
75
|
+
};
|
|
76
|
+
}
|
|
77
|
+
const runDir = path.join(rootDir, '.ai-engineering-loop', 'runs', run.runId);
|
|
78
|
+
fs.writeFileSync(path.join(runDir, 'goal-contract.json'), `${JSON.stringify(quickGoal(run.runId, task), null, 2)}\n`, { mode: 0o600 });
|
|
79
|
+
const gate = invoke(rootDir, ['gate', 'goal', '--run', run.runId]);
|
|
80
|
+
return {
|
|
81
|
+
ok: true,
|
|
82
|
+
runId: run.runId,
|
|
83
|
+
tier: 'T0',
|
|
84
|
+
quick: true,
|
|
85
|
+
state: gate.state,
|
|
86
|
+
learnings: run.learnings || [],
|
|
87
|
+
next: `Make the change, then: npx ai-engineering-loop quick finish --run ${run.runId}`
|
|
88
|
+
};
|
|
89
|
+
}
|
|
90
|
+
|
|
91
|
+
function writeClaimedVsReality(rootDir, { changedPaths, commands }) {
|
|
92
|
+
const reality = commands
|
|
93
|
+
.map((c) => `\`${oneLine(c.command, 120)}\` exit ${c.exitCode}${c.testCounts ? ` (${c.testCounts.passed ?? '?'} passed, ${c.testCounts.failed ?? 0} failed)` : ''}`)
|
|
94
|
+
.join('; ');
|
|
95
|
+
const claimed = `Changed ${changedPaths.length} file(s): ${changedPaths.slice(0, 10).join(', ')}${changedPaths.length > 10 ? ', …' : ''}`;
|
|
96
|
+
const text = [
|
|
97
|
+
'# Claimed vs Reality',
|
|
98
|
+
'',
|
|
99
|
+
'> Written by `quick finish` from the gated diff and the recorded verification output.',
|
|
100
|
+
'',
|
|
101
|
+
'| AC | Claimed | Reality from command log |',
|
|
102
|
+
'|---|---|---|',
|
|
103
|
+
`| AC-1 | ${oneLine(claimed, 600)} | ${oneLine(reality, 600)} |`,
|
|
104
|
+
''
|
|
105
|
+
].join('\n');
|
|
106
|
+
const filePath = path.join(rootDir, CLAIMED_PATH);
|
|
107
|
+
fs.mkdirSync(path.dirname(filePath), { recursive: true, mode: 0o700 });
|
|
108
|
+
fs.writeFileSync(filePath, text, { mode: 0o600 });
|
|
109
|
+
return CLAIMED_PATH;
|
|
110
|
+
}
|
|
111
|
+
|
|
112
|
+
function finishQuick(rootDir, { runId, commands = [], shell = false, serial = false, timeoutMs = null } = {}) {
|
|
113
|
+
const runArgs = runId ? ['--run', runId] : [];
|
|
114
|
+
const maker = invoke(rootDir, ['gate', 'maker', '--from-git', ...runArgs]);
|
|
115
|
+
const id = maker.runId;
|
|
116
|
+
if (maker.tier && maker.tier !== 'T0') {
|
|
117
|
+
return {
|
|
118
|
+
ok: false,
|
|
119
|
+
code: 'TIER_RAISED',
|
|
120
|
+
runId: id,
|
|
121
|
+
tier: maker.tier,
|
|
122
|
+
state: maker.state,
|
|
123
|
+
next: `The diff raised the tier to ${maker.tier}. Continue the normal loop: verify --run ${id}, then the Devil's Advocate review.`
|
|
124
|
+
};
|
|
125
|
+
}
|
|
126
|
+
const verify = invoke(rootDir, [
|
|
127
|
+
'verify', '--run', id,
|
|
128
|
+
...commands.flatMap((command) => ['--cmd', command]),
|
|
129
|
+
...(shell ? ['--shell'] : []),
|
|
130
|
+
...(serial ? ['--serial'] : []),
|
|
131
|
+
...(timeoutMs ? ['--timeout-ms', String(timeoutMs)] : [])
|
|
132
|
+
]);
|
|
133
|
+
const diffPath = path.join(rootDir, '.ai-engineering-loop', 'runs', id, 'diff.patch');
|
|
134
|
+
const changedPaths = [...new Set(extractChangedPaths(fs.readFileSync(diffPath, 'utf8')))];
|
|
135
|
+
const claimed = writeClaimedVsReality(rootDir, { changedPaths, commands: verify.commands.map((c) => ({ ...c, testCounts: verify.testCounts })) });
|
|
136
|
+
invoke(rootDir, ['gate', 'review', '--run', id]);
|
|
137
|
+
const judge = invoke(rootDir, ['gate', 'judge', '--run', id]);
|
|
138
|
+
return {
|
|
139
|
+
ok: true,
|
|
140
|
+
runId: id,
|
|
141
|
+
tier: 'T0',
|
|
142
|
+
state: judge.state,
|
|
143
|
+
commands: verify.commands,
|
|
144
|
+
changedPaths,
|
|
145
|
+
claimedVsReality: claimed,
|
|
146
|
+
next: judge.state === 'JUDGED_PASS'
|
|
147
|
+
? `After the human approves: npx ai-engineering-loop deliver --run ${id} --destination <where> --summary "<text>" --approve --by <name>`
|
|
148
|
+
: `Judge returned ${judge.state}; fix and run quick finish again.`
|
|
149
|
+
};
|
|
150
|
+
}
|
|
151
|
+
|
|
152
|
+
module.exports = { startQuick, finishQuick, quickGoal };
|
package/lib/receipt.js
CHANGED
|
@@ -6,10 +6,10 @@
|
|
|
6
6
|
// `verify-receipt` re-checks all of it, so a reviewer (or CI) can trust the
|
|
7
7
|
// PR body without re-reading the Run.
|
|
8
8
|
|
|
9
|
-
const crypto = require('crypto');
|
|
10
9
|
const fs = require('fs');
|
|
11
10
|
const path = require('path');
|
|
12
11
|
const { execFileSync } = require('child_process');
|
|
12
|
+
const { sha256: hashBytes } = require('./fsutil.js');
|
|
13
13
|
const { getCurrentRun, loadRun, runsDir } = require('./run-state.js');
|
|
14
14
|
const { canonicalJson, sha256 } = require('./recipe.js');
|
|
15
15
|
const { parseClaimedVsReality } = require('./claimed-vs-reality.js');
|
|
@@ -45,10 +45,6 @@ function resolveRun(rootDir, runId) {
|
|
|
45
45
|
return current;
|
|
46
46
|
}
|
|
47
47
|
|
|
48
|
-
function hashBytes(buffer) {
|
|
49
|
-
return crypto.createHash('sha256').update(buffer).digest('hex');
|
|
50
|
-
}
|
|
51
|
-
|
|
52
48
|
function repoRelative(rootDir, filePath) {
|
|
53
49
|
return path.relative(rootDir, filePath).split(path.sep).join('/');
|
|
54
50
|
}
|
package/lib/recipe-builder.js
CHANGED
|
@@ -2,14 +2,10 @@
|
|
|
2
2
|
|
|
3
3
|
const fs = require('fs');
|
|
4
4
|
const path = require('path');
|
|
5
|
-
const crypto = require('crypto');
|
|
6
5
|
const { NODE_TYPES } = require('./node-registry.js');
|
|
7
6
|
const { ensurePrivateRuntimeIgnores } = require('./runtime-files.js');
|
|
8
|
-
const {
|
|
9
|
-
|
|
10
|
-
migrateLegacyRecipe,
|
|
11
|
-
planGraphMerge
|
|
12
|
-
} = require('./recipe-graph.js');
|
|
7
|
+
const { writePrivateJson } = require('./fsutil.js');
|
|
8
|
+
const { migrateLegacyRecipe } = require('./recipe-graph.js');
|
|
13
9
|
const {
|
|
14
10
|
RECIPE_ID,
|
|
15
11
|
canonicalJson,
|
|
@@ -48,14 +44,7 @@ function atomicWriteJson(filePath, value, { replace = false } = {}) {
|
|
|
48
44
|
if (!replace && fs.existsSync(filePath)) {
|
|
49
45
|
throw builderError(`Recipe already exists: ${path.basename(filePath, '.json')}`, 'RECIPE_EXISTS');
|
|
50
46
|
}
|
|
51
|
-
|
|
52
|
-
const temporary = `${filePath}.${process.pid}.${crypto.randomBytes(4).toString('hex')}.tmp`;
|
|
53
|
-
try {
|
|
54
|
-
fs.writeFileSync(temporary, `${JSON.stringify(value, null, 2)}\n`, { mode: 0o600 });
|
|
55
|
-
fs.renameSync(temporary, filePath);
|
|
56
|
-
} finally {
|
|
57
|
-
try { fs.unlinkSync(temporary); } catch {}
|
|
58
|
-
}
|
|
47
|
+
writePrivateJson(filePath, value);
|
|
59
48
|
}
|
|
60
49
|
|
|
61
50
|
function assertProjectId(rootDir, id) {
|
|
@@ -71,25 +60,6 @@ function validateCandidate(recipe) {
|
|
|
71
60
|
return result;
|
|
72
61
|
}
|
|
73
62
|
|
|
74
|
-
function planRecipeImport(existingRecipe, importedRecipe, operation) {
|
|
75
|
-
if (!['replace', 'merge', 'cancel'].includes(operation)) {
|
|
76
|
-
throw builderError('Import operation must be replace, merge, or cancel', 'INVALID_IMPORT_OPERATION');
|
|
77
|
-
}
|
|
78
|
-
validateCandidate(existingRecipe);
|
|
79
|
-
validateCandidate(importedRecipe);
|
|
80
|
-
if (operation !== 'merge') return { operation, nodeIdMap: {}, edgeIdMap: {} };
|
|
81
|
-
return { operation, ...planGraphMerge(existingRecipe, importedRecipe) };
|
|
82
|
-
}
|
|
83
|
-
|
|
84
|
-
function applyRecipeImport(existingRecipe, importedRecipe, plan) {
|
|
85
|
-
if (!plan || !['replace', 'merge', 'cancel'].includes(plan.operation)) {
|
|
86
|
-
throw builderError('A valid import preview is required', 'INVALID_IMPORT_PLAN');
|
|
87
|
-
}
|
|
88
|
-
if (plan.operation === 'cancel') return existingRecipe;
|
|
89
|
-
if (plan.operation === 'replace') return migrateLegacyRecipe(importedRecipe);
|
|
90
|
-
return mergeGraphs(existingRecipe, importedRecipe, plan);
|
|
91
|
-
}
|
|
92
|
-
|
|
93
63
|
function cloneRecipe(rootDir, sourceId, targetId, { description } = {}) {
|
|
94
64
|
assertProjectId(rootDir, targetId);
|
|
95
65
|
const source = loadRecipe(rootDir, sourceId).recipe;
|
|
@@ -217,10 +187,8 @@ function diffRecipes(rootDir, leftId, rightId) {
|
|
|
217
187
|
}
|
|
218
188
|
|
|
219
189
|
module.exports = {
|
|
220
|
-
applyRecipeImport,
|
|
221
190
|
cloneRecipe,
|
|
222
191
|
installRecipe,
|
|
223
|
-
planRecipeImport,
|
|
224
192
|
recipeCatalog,
|
|
225
193
|
inspectRecipe,
|
|
226
194
|
diffRecipes
|