ai-engineering-loop 2.6.0 → 2.8.0

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 (53) hide show
  1. package/.agents/skills/ael/SKILL.md +1 -1
  2. package/.agents/skills/ael/references/loop-commands.md +13 -1
  3. package/.agents/skills/ael/references/loop-deliver.md +1 -1
  4. package/.agents/skills/ael/references/loop-goal.md +2 -0
  5. package/.agents/skills/ael/references/loop-maker.md +1 -6
  6. package/.agents/skills/ael/references/loop-report.md +8 -0
  7. package/.agents/skills/ael/references/loop-verify.md +11 -9
  8. package/.claude/agents/devil-advocate.md +7 -2
  9. package/.claude/agents/judge.md +3 -1
  10. package/.grok/agents/devil-advocate.md +7 -2
  11. package/.grok/agents/judge.md +3 -1
  12. package/CHANGELOG.md +31 -0
  13. package/agents/devil-advocate.md +8 -0
  14. package/agents/judge.md +1 -0
  15. package/agents/shared/devil-advocate.body.md +7 -2
  16. package/agents/shared/judge.body.md +3 -1
  17. package/bin/ai-engineering-loop.js +326 -21
  18. package/core/bench.md +15 -0
  19. package/core/formal-statechart.md +49 -0
  20. package/core/goal-contract.md +53 -0
  21. package/core/verification-loop.md +106 -1
  22. package/lib/ablation.js +52 -0
  23. package/lib/ac-status.js +95 -0
  24. package/lib/baseline.js +138 -0
  25. package/lib/bench.js +24 -1
  26. package/lib/budget.js +9 -1
  27. package/lib/calibration.js +129 -0
  28. package/lib/claimed-vs-reality.js +106 -12
  29. package/lib/clean-state.js +85 -0
  30. package/lib/cli-args.js +5 -3
  31. package/lib/comment-findings.js +111 -0
  32. package/lib/debt.js +110 -0
  33. package/lib/feature-map.js +57 -0
  34. package/lib/features.js +95 -0
  35. package/lib/flow.js +86 -0
  36. package/lib/gates.js +257 -41
  37. package/lib/harness-audit.js +141 -0
  38. package/lib/learn.js +24 -1
  39. package/lib/no-progress.js +6 -1
  40. package/lib/receipt.js +33 -2
  41. package/lib/reverts.js +77 -0
  42. package/lib/review-ingest.js +54 -2
  43. package/lib/rule-check.js +90 -0
  44. package/lib/rules.js +205 -0
  45. package/lib/stats.js +110 -0
  46. package/lib/sync-hosts.js +1 -0
  47. package/lib/verification-integrity.js +164 -0
  48. package/lib/verification-lanes.js +96 -0
  49. package/lib/verification-recorder.js +20 -8
  50. package/package.json +1 -1
  51. package/schemas/finding-ledger.schema.json +14 -0
  52. package/schemas/goal-contract.schema.json +50 -0
  53. package/schemas/verdict.schema.json +2 -1
@@ -22,6 +22,20 @@
22
22
  },
23
23
  "additionalProperties": true
24
24
  }
25
+ },
26
+ "rubricScores": {
27
+ "description": "C-09: one score per rubric criterion of a subjective AC",
28
+ "type": "array",
29
+ "items": {
30
+ "type": "object",
31
+ "required": ["ac", "criterion", "pass", "evidence"],
32
+ "properties": {
33
+ "ac": { "type": "string", "pattern": "^AC-[0-9]+$" },
34
+ "criterion": { "type": "string", "minLength": 1 },
35
+ "pass": { "type": "boolean" },
36
+ "evidence": { "type": "string", "minLength": 1 }
37
+ }
38
+ }
25
39
  }
26
40
  },
27
41
  "additionalProperties": true
@@ -52,6 +52,37 @@
52
52
  "type": "string",
53
53
  "minLength": 1
54
54
  }
55
+ },
56
+ "rubric": {
57
+ "description": "Optional (C-09) for subjective ACs: criteria the Devil's Advocate scores with evidence in ledger.rubricScores. Judge PASS needs every criterion scored pass:true.",
58
+ "type": "array",
59
+ "items": {
60
+ "type": "object",
61
+ "required": ["criterion", "pass"],
62
+ "properties": {
63
+ "criterion": { "type": "string", "minLength": 1 },
64
+ "pass": { "type": "string", "minLength": 1, "description": "What counts as passing, e.g. 'contrast >= 4.5:1 (WCAG AA)'" }
65
+ }
66
+ }
67
+ },
68
+ "verify": {
69
+ "description": "Optional (C-07, C-23). What proves this AC: an exact `command`, a verification catalog `key` (e.g. pentest, property, formal; default verify runs it when configured), or a `layer` (any passing command of that layer). Gate judge refuses PASS until it passed on the gated diff.",
70
+ "type": "object",
71
+ "anyOf": [
72
+ { "required": ["command"] },
73
+ { "required": ["key"] },
74
+ { "required": ["layer"] }
75
+ ],
76
+ "properties": {
77
+ "command": { "type": "string", "minLength": 1 },
78
+ "key": { "enum": ["build", "typecheck", "lint", "security", "test_unit", "test_related", "test_all", "property", "fuzz", "mutation", "coverage", "benchmark", "smoke", "contract", "e2e", "pentest", "formal"] },
79
+ "layer": { "enum": ["L1", "L2", "L3", "LF"] }
80
+ },
81
+ "additionalProperties": false
82
+ },
83
+ "crossComponent": {
84
+ "description": "Optional (#110). true when the AC crosses a component boundary; claimed-vs-reality then needs L3 (end-to-end) evidence or an 'L3 waived: <reason>' Layer cell before gate review.",
85
+ "type": "boolean"
55
86
  }
56
87
  },
57
88
  "additionalProperties": true
@@ -104,6 +135,25 @@
104
135
  },
105
136
  "additionalProperties": false
106
137
  },
138
+ "flow": {
139
+ "description": "Optional (C-21): the intended user flow as transitions. goal draft renders it as a From | Event | To table for the human to check before freeze; the receipt shows it next to the verification artifacts.",
140
+ "type": "array",
141
+ "minItems": 1,
142
+ "items": {
143
+ "type": "object",
144
+ "required": ["from", "event", "to"],
145
+ "properties": {
146
+ "from": { "type": "string", "minLength": 1 },
147
+ "event": { "type": "string", "minLength": 1 },
148
+ "to": { "type": "string", "minLength": 1 }
149
+ }
150
+ }
151
+ },
152
+ "featureId": {
153
+ "description": "Optional (C-15): id of the feature in .ai-engineering-loop/features.json this Goal delivers. Goal freeze marks it active; delivery marks it passing with this Run as evidence.",
154
+ "type": "string",
155
+ "minLength": 1
156
+ },
107
157
  "outOfScope": {
108
158
  "description": "Optional list of things this Goal must not change.",
109
159
  "type": "array",
@@ -11,7 +11,8 @@
11
11
  "reason": { "type": "string", "minLength": 1 },
12
12
  "action": { "type": "string", "minLength": 1 },
13
13
  "modelTier": { "enum": ["CHEAP_MODEL", "STRONG_MODEL", "HUMAN"] },
14
- "humanApproved": { "type": "boolean" }
14
+ "humanApproved": { "type": "boolean" },
15
+ "cause": { "description": "C-12: harness layer a non-PASS verdict points at", "enum": ["task_spec", "context", "environment", "verification", "state", "model"] }
15
16
  },
16
17
  "additionalProperties": true
17
18
  }