@cassiomc1/forgeloop 1.13.0 → 1.14.0
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/AGENT_COMPATIBILITY.md +8 -0
- package/DOCS_INDEX.md +35 -3
- package/ENG/nodejs-backend-development-eng.md +2 -2
- package/ENG/sec-code-eng.md +7 -7
- package/EXECUTION_STATE.md +12 -0
- package/LOOP_ENGINEERING.md +28 -2
- package/ORCHESTRATOR_INTEGRATION.md +9 -5
- package/PROTOCOL_INTEGRATION.md +55 -2
- package/README.md +40 -25
- package/TERMINOLOGY.md +2 -0
- package/THREAT_MODEL.md +140 -1
- package/completions/_forgeloop +19 -1
- package/completions/forgeloop.bash +37 -1
- package/completions/forgeloop.fish +123 -1
- package/docs/ADVISORY_CONTEXT.md +25 -0
- package/docs/AGENT_BROWSER_ADAPTER.md +81 -0
- package/docs/AGENT_BROWSER_VERIFICATION.md +6 -0
- package/docs/AGENT_PROTOCOL_SUMMARY.md +27 -2
- package/docs/AGENT_SKILL.md +66 -0
- package/docs/ARTIFACT_REFERENCE.md +123 -0
- package/docs/AUDIT_UX.md +46 -0
- package/docs/BROWSER_VERIFICATION.md +136 -0
- package/docs/CLI_REFERENCE.md +366 -6
- package/docs/CODE_ATTESTATION.md +2 -2
- package/docs/DOCUMENTATION_GUIDE.md +32 -11
- package/docs/JEV_BENCHMARKS.md +31 -0
- package/docs/MODEL_ROUTING.md +37 -0
- package/docs/OPENSRC_ADAPTER.md +241 -0
- package/docs/PACKAGE_CONTENTS.md +35 -8
- package/docs/PROVIDERS.md +126 -0
- package/docs/PROVIDER_ARCHITECTURE.md +199 -0
- package/docs/RECIPES.md +9 -0
- package/docs/RELEASE_CHECKLIST.md +38 -5
- package/docs/SECURITY_REVIEW.md +71 -0
- package/docs/SEMANTIC_DECISION_PLANE.md +71 -0
- package/docs/TEST_INTELLIGENCE.md +29 -0
- package/docs/TEST_PRUNING.md +14 -0
- package/docs/TROUBLESHOOTING.md +198 -1
- package/docs/UNIVERSAL_INTEGRATION.md +31 -0
- package/docs/assets/diagrams/forgeloop-code-attestation-flow.html +2 -2
- package/docs/assets/diagrams/forgeloop-code-attestation-flow.receipt.json +5 -5
- package/docs/assets/diagrams/forgeloop-code-attestation-flow.svg +1 -1
- package/docs/assets/diagrams/forgeloop-engineering-flow.html +39 -26
- package/docs/assets/diagrams/forgeloop-engineering-flow.receipt.json +6 -6
- package/docs/assets/diagrams/forgeloop-engineering-flow.svg +26 -26
- package/docs/assets/diagrams/forgeloop-verification-trust-flow.html +2 -1
- package/docs/assets/diagrams/forgeloop-verification-trust-flow.receipt.json +5 -5
- package/docs/assets/diagrams/forgeloop-verification-trust-flow.svg +1 -1
- package/docs/diagrams/README.md +13 -9
- package/docs/diagrams/forgeloop-code-attestation-flow.workflow.json +1 -1
- package/docs/diagrams/forgeloop-engineering-flow.workflow.json +24 -19
- package/docs/diagrams/forgeloop-verification-trust-flow.workflow.json +1 -0
- package/docs/diagrams/reviews/forgeloop-code-attestation-flow.review.json +4 -4
- package/docs/diagrams/reviews/forgeloop-engineering-flow.review.json +4 -4
- package/docs/diagrams/reviews/forgeloop-verification-trust-flow.review.json +4 -4
- package/docs/documentation-manifest.json +750 -5
- package/docs/protocol-requirements.json +24 -0
- package/package.json +30 -3
- package/schemas/config.schema.json +14 -0
- package/schemas/context-plan.schema.json +18 -0
- package/schemas/semantic-decision.schema.json +46 -0
- package/schemas/test-utility.schema.json +44 -0
- package/scripts/CI_VALIDATORS.md +6 -6
- package/scripts/benchmark-jev.mjs +5 -0
- package/scripts/benchmark-test-intelligence.mjs +4 -0
- package/scripts/generate-agent-protocol-summary.mjs +4 -1
- package/scripts/generate-forgeloop-skill.mjs +133 -0
- package/scripts/jev-smoke.mjs +19 -0
- package/skills/forgeloop/README.md +9 -0
- package/skills/forgeloop/SKILL.md +77 -0
- package/skills/forgeloop/references/lifecycle.md +9 -0
- package/skills/forgeloop/references/recovery.md +7 -0
- package/skills/forgeloop/references/verification.md +7 -0
- package/src/adapters/agent-browser/assertions.js +47 -0
- package/src/adapters/agent-browser/commands.js +54 -0
- package/src/adapters/agent-browser/index.js +3 -0
- package/src/adapters/agent-browser/locator.js +40 -0
- package/src/adapters/agent-browser/process.js +215 -0
- package/src/adapters/agent-browser/provider.js +313 -0
- package/src/adapters/emulated-services/constants.js +24 -0
- package/src/adapters/emulated-services/index.js +7 -0
- package/src/adapters/emulated-services/process.js +162 -0
- package/src/adapters/emulated-services/provider.js +282 -0
- package/src/adapters/opensrc/normalize.js +90 -0
- package/src/adapters/opensrc/process.js +248 -0
- package/src/adapters/opensrc/provider.js +338 -0
- package/src/adapters/opensrc/search.js +264 -0
- package/src/adapters/typesafe/client.js +28 -0
- package/src/adapters/typesafe/engine.js +63 -0
- package/src/adapters/typesafe/normalize.js +41 -0
- package/src/cli.js +108 -0
- package/src/commands/checkpoint-revalidate.js +176 -0
- package/src/commands/context-plan.js +38 -0
- package/src/commands/contract-create.js +264 -0
- package/src/commands/contract-revise.js +236 -0
- package/src/commands/decision-show.js +14 -0
- package/src/commands/decision-status.js +22 -0
- package/src/commands/discover.js +41 -0
- package/src/commands/doctor.js +15 -0
- package/src/commands/gate-record.js +205 -0
- package/src/commands/gate-revalidate.js +137 -0
- package/src/commands/model-route.js +32 -0
- package/src/commands/route.js +146 -18
- package/src/commands/semantic-plan.js +17 -0
- package/src/commands/task-abandon.js +224 -0
- package/src/commands/task-migrate-contract-bootstrap-repair.js +288 -0
- package/src/commands/task-repair-contract-bootstrap.js +263 -0
- package/src/commands/test-inventory.js +5 -0
- package/src/commands/test-prune-plan.js +5 -0
- package/src/commands/test-prune-probe.js +5 -0
- package/src/commands/test-utility.js +5 -0
- package/src/commands/validate-protocol.js +10 -1
- package/src/core/artifact-registry.js +24 -0
- package/src/core/audit-ux.js +514 -0
- package/src/core/browser-verification/constants.js +149 -0
- package/src/core/browser-verification/normalize.js +254 -0
- package/src/core/browser-verification/provider.js +519 -0
- package/src/core/browser-verification/service.js +115 -0
- package/src/core/checkpoint-revalidation.js +319 -0
- package/src/core/cli-command-definitions.js +241 -0
- package/src/core/command-executors.js +110 -0
- package/src/core/command-input.js +115 -43
- package/src/core/completion-artifacts.js +14 -5
- package/src/core/completion.js +4 -6
- package/src/core/config.js +3 -0
- package/src/core/context-compiler/budget.js +9 -0
- package/src/core/context-compiler/candidates.js +39 -0
- package/src/core/context-compiler/compiler.js +63 -0
- package/src/core/context-compiler/fingerprint.js +11 -0
- package/src/core/context-compiler/policy.js +13 -0
- package/src/core/context-compiler/result.js +23 -0
- package/src/core/contract-bootstrap-recovery.js +655 -0
- package/src/core/contract-revision.js +210 -0
- package/src/core/decision/artifact.js +69 -0
- package/src/core/decision/benchmarks.js +103 -0
- package/src/core/decision/cache.js +27 -0
- package/src/core/decision/constants.js +58 -0
- package/src/core/decision/cutover.js +34 -0
- package/src/core/decision/engine.js +22 -0
- package/src/core/decision/errors.js +68 -0
- package/src/core/decision/events.js +101 -0
- package/src/core/decision/freshness.js +19 -0
- package/src/core/decision/normalizers/index.js +115 -0
- package/src/core/decision/policy.js +18 -0
- package/src/core/decision/projection.js +16 -0
- package/src/core/decision/question-registry.js +201 -0
- package/src/core/decision/request.js +26 -0
- package/src/core/decision/resolver.js +130 -0
- package/src/core/decision/result.js +58 -0
- package/src/core/decision/service.js +156 -0
- package/src/core/decision/state-builder.js +65 -0
- package/src/core/decision/task-bindings.js +30 -0
- package/src/core/decision/test-provider.js +32 -0
- package/src/core/decision/thresholds.js +15 -0
- package/src/core/error-codes.js +278 -0
- package/src/core/events.js +226 -57
- package/src/core/evidence-readiness.js +9 -0
- package/src/core/execution-prerequisites.js +14 -0
- package/src/core/execution-profile.js +63 -38
- package/src/core/gate-provenance.js +124 -0
- package/src/core/integration-invocation-policy.js +27 -4
- package/src/core/integration-resources.js +86 -61
- package/src/core/model-router/constants.js +10 -0
- package/src/core/model-router/policy.js +103 -0
- package/src/core/model-router/router.js +37 -0
- package/src/core/next-action-model.js +58 -0
- package/src/core/next-action-phases.js +130 -42
- package/src/core/next-action-refresh.js +43 -9
- package/src/core/next-action-review-phase.js +7 -2
- package/src/core/next-action.js +35 -7
- package/src/core/phase.js +128 -10
- package/src/core/preflight-consistency.js +23 -9
- package/src/core/preflight-loaders.js +37 -5
- package/src/core/protocol-info.js +65 -0
- package/src/core/protocol.js +20 -0
- package/src/core/reconcile-closure.js +128 -52
- package/src/core/recovery-history.js +1 -0
- package/src/core/resumability.js +154 -44
- package/src/core/route-artifact.js +15 -1
- package/src/core/router.js +67 -1
- package/src/core/runtime-context.js +118 -61
- package/src/core/schema-validation.js +3 -0
- package/src/core/security-review/constants.js +64 -0
- package/src/core/security-review/normalize.js +245 -0
- package/src/core/security-review/provider.js +204 -0
- package/src/core/security-review/service.js +134 -0
- package/src/core/semantic-planning/constants.js +19 -0
- package/src/core/semantic-planning/projection.js +94 -0
- package/src/core/semantic-planning/service.js +15 -0
- package/src/core/sources.js +37 -0
- package/src/core/task-claim-state.js +201 -1
- package/src/core/task-conflict-inspection.js +31 -5
- package/src/core/task-paths.js +13 -0
- package/src/core/task-recovery.js +1 -0
- package/src/core/templates.js +3 -0
- package/src/core/test-intelligence/benchmarks.js +68 -0
- package/src/core/test-intelligence/inventory.js +73 -0
- package/src/core/test-intelligence/prune.js +90 -0
- package/src/core/test-intelligence/semantic-state.js +15 -0
- package/src/core/test-intelligence/service.js +40 -0
- package/src/core/test-intelligence/utility.js +50 -0
- package/src/core/trace.js +11 -7
- package/src/core/transaction.js +1 -0
- package/src/integration.d.ts +492 -0
- package/src/integration.js +54 -0
- package/src/providers/README.md +47 -0
- package/src/providers/capabilities.js +46 -0
- package/src/providers/errors.js +15 -0
- package/src/providers/index.js +29 -0
- package/src/providers/json-snapshot.js +105 -0
- package/src/providers/registry.js +152 -0
package/docs/CLI_REFERENCE.md
CHANGED
|
@@ -18,6 +18,135 @@ Commands that support structured machine-readable output document `--json` in th
|
|
|
18
18
|
|
|
19
19
|
<!-- END FORGELOOP GENERATED: cli-common-options -->
|
|
20
20
|
|
|
21
|
+
## Semantic decision projections
|
|
22
|
+
|
|
23
|
+
### `decision-status`
|
|
24
|
+
|
|
25
|
+
Reports the pinned Jev configuration and, only when explicitly requested,
|
|
26
|
+
performs a bounded provider health check.
|
|
27
|
+
|
|
28
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:decision-status:options -->
|
|
29
|
+
|
|
30
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
31
|
+
- `--health`: perform a bounded live Jev health check when credentials are configured
|
|
32
|
+
- `--json`: emit decision-plane status as JSON
|
|
33
|
+
|
|
34
|
+
<!-- END FORGELOOP GENERATED: cli:decision-status:options -->
|
|
35
|
+
|
|
36
|
+
### `decision-show`
|
|
37
|
+
|
|
38
|
+
Shows a persisted semantic decision without performing a live request.
|
|
39
|
+
|
|
40
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:decision-show:options -->
|
|
41
|
+
|
|
42
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
43
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
44
|
+
- `--decision <id>`: decision artifact ID
|
|
45
|
+
- `--json`: emit the decision artifact as JSON
|
|
46
|
+
|
|
47
|
+
<!-- END FORGELOOP GENERATED: cli:decision-show:options -->
|
|
48
|
+
|
|
49
|
+
### `context-plan`
|
|
50
|
+
|
|
51
|
+
Compiles a bounded, non-authoritative context plan.
|
|
52
|
+
|
|
53
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:context-plan:options -->
|
|
54
|
+
|
|
55
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
56
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
57
|
+
- `--decision <id>`: persisted Jev decision artifact ID
|
|
58
|
+
- `--profile <profile>`: bounded context budget profile
|
|
59
|
+
- `--json`: emit the bounded context plan as JSON
|
|
60
|
+
|
|
61
|
+
<!-- END FORGELOOP GENERATED: cli:context-plan:options -->
|
|
62
|
+
|
|
63
|
+
### `model-route`
|
|
64
|
+
|
|
65
|
+
Projects the deterministic model-routing floor. Jev may escalate but cannot
|
|
66
|
+
lower the floor or select vendor-specific models.
|
|
67
|
+
|
|
68
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:model-route:options -->
|
|
69
|
+
|
|
70
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
71
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
72
|
+
- `--decision <id>`: persisted Jev decision artifact ID
|
|
73
|
+
- `--work <type>`: declared work type
|
|
74
|
+
- `--surface <value>`: affected surface (repeatable)
|
|
75
|
+
- `--risk <value>`: task risk (repeatable)
|
|
76
|
+
- `--platform <value>`: affected platform (repeatable)
|
|
77
|
+
- `--behavior-change`: declare behavior change
|
|
78
|
+
- `--executable-change`: declare executable/configuration change
|
|
79
|
+
- `--generation-required`: declare that generation is required
|
|
80
|
+
- `--architecture-change`: declare an architectural change
|
|
81
|
+
- `--ambiguous`: declare unresolved ambiguity
|
|
82
|
+
- `--json`: emit model-route projection as JSON
|
|
83
|
+
|
|
84
|
+
<!-- END FORGELOOP GENERATED: cli:model-route:options -->
|
|
85
|
+
|
|
86
|
+
### `semantic-plan`
|
|
87
|
+
|
|
88
|
+
Projects bounded failure, diagnosis, or review planning from ForgeLoop-owned
|
|
89
|
+
question-set categories.
|
|
90
|
+
|
|
91
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:semantic-plan:options -->
|
|
92
|
+
|
|
93
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
94
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
95
|
+
- `--decision <id>`: persisted Jev decision artifact ID
|
|
96
|
+
- `--kind <failure|diagnosis|review>`: bounded semantic planning projection
|
|
97
|
+
- `--input <json>`: bounded failure, diagnosis, or review context
|
|
98
|
+
- `--json`: emit semantic plan projection as JSON
|
|
99
|
+
|
|
100
|
+
<!-- END FORGELOOP GENERATED: cli:semantic-plan:options -->
|
|
101
|
+
|
|
102
|
+
### `test-inventory`
|
|
103
|
+
|
|
104
|
+
Discovers deterministic test units and stable test IDs.
|
|
105
|
+
|
|
106
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:test-inventory:options -->
|
|
107
|
+
|
|
108
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
109
|
+
- `--json`: emit deterministic test inventory as JSON
|
|
110
|
+
|
|
111
|
+
<!-- END FORGELOOP GENERATED: cli:test-inventory:options -->
|
|
112
|
+
|
|
113
|
+
### `test-utility`
|
|
114
|
+
|
|
115
|
+
Persists non-evidence test utility analysis. It never removes tests.
|
|
116
|
+
|
|
117
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:test-utility:options -->
|
|
118
|
+
|
|
119
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
120
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
121
|
+
- `--json`: emit test utility analysis as JSON
|
|
122
|
+
|
|
123
|
+
<!-- END FORGELOOP GENERATED: cli:test-utility:options -->
|
|
124
|
+
|
|
125
|
+
### `test-prune-plan`
|
|
126
|
+
|
|
127
|
+
Projects a safe non-destructive test pruning plan.
|
|
128
|
+
|
|
129
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:test-prune-plan:options -->
|
|
130
|
+
|
|
131
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
132
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
133
|
+
- `--json`: emit test prune plan as JSON
|
|
134
|
+
|
|
135
|
+
<!-- END FORGELOOP GENERATED: cli:test-prune-plan:options -->
|
|
136
|
+
|
|
137
|
+
### `test-prune-probe`
|
|
138
|
+
|
|
139
|
+
Probes only eligible candidates in isolation and never modifies the live tree.
|
|
140
|
+
|
|
141
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:test-prune-probe:options -->
|
|
142
|
+
|
|
143
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
144
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
145
|
+
- `--test <id>`: stable test ID to probe in isolation
|
|
146
|
+
- `--json`: emit test prune probe as JSON
|
|
147
|
+
|
|
148
|
+
<!-- END FORGELOOP GENERATED: cli:test-prune-probe:options -->
|
|
149
|
+
|
|
21
150
|
---
|
|
22
151
|
|
|
23
152
|
## CLI Syntax Contract
|
|
@@ -58,10 +187,10 @@ error codes. Default output and default JSON remain unchanged.
|
|
|
58
187
|
|
|
59
188
|
| Category | Commands |
|
|
60
189
|
| --- | --- |
|
|
61
|
-
| **Inspection & Diagnostics** | [`protocol-info`](#protocol-info), [`doctor`](#doctor), [`index-status`](#index-status), [`search`](#search), [`metrics`](#metrics), [`usage-record`](#usage-record), [`efficiency`](#efficiency), [`eval`](#eval), [`history`](#history), [`trace`](#trace), [`reflect`](#reflect), [`progress`](#progress), [`profile-interview`](#profile-interview), [`inspect`](#inspect), [`status`](#status), [`validate-state`](#validate-state), [`validate-protocol`](#validate-protocol) |
|
|
62
|
-
| **
|
|
63
|
-
| **Lifecycle & State** | [`activate`](#activate), [`route`](#route), [`preflight`](#preflight), [`advance`](#advance), [`next`](#next), [`record-diagnosis`](#record-diagnosis), [`record-intervention`](#record-intervention), [`record-hypothesis-disposition`](#record-hypothesis-disposition), [`record-decision-criterion`](#record-decision-criterion), [`complete`](#complete), [`clear-state`](#clear-state), [`reconcile-closure`](#reconcile-closure), [`task-create`](#task-create), [`task-list`](#task-list), [`task-show`](#task-show), [`task-lock-status`](#task-lock-status), [`task-scope`](#task-scope) |
|
|
64
|
-
| **
|
|
190
|
+
| **Inspection & Diagnostics** | [`protocol-info`](#protocol-info), [`decision-status`](#decision-status), [`decision-show`](#decision-show), [`context-plan`](#context-plan), [`model-route`](#model-route), [`semantic-plan`](#semantic-plan), [`test-inventory`](#test-inventory), [`test-utility`](#test-utility), [`test-prune-plan`](#test-prune-plan), [`doctor`](#doctor), [`index-status`](#index-status), [`search`](#search), [`metrics`](#metrics), [`usage-record`](#usage-record), [`efficiency`](#efficiency), [`eval`](#eval), [`history`](#history), [`trace`](#trace), [`reflect`](#reflect), [`progress`](#progress), [`profile-interview`](#profile-interview), [`inspect`](#inspect), [`status`](#status), [`validate-state`](#validate-state), [`validate-protocol`](#validate-protocol) |
|
|
191
|
+
| **Verification & Completion** | [`test-prune-probe`](#test-prune-probe), [`quality-baseline`](#quality-baseline), [`quality-verify`](#quality-verify), [`quality-status`](#quality-status), [`prepare-completion`](#prepare-completion), [`run-check`](#run-check), [`record-check`](#record-check), [`record-terminal-result`](#record-terminal-result), [`audit`](#audit), [`report`](#report), [`validate-receipt`](#validate-receipt), [`verify-scope`](#verify-scope) |
|
|
192
|
+
| **Lifecycle & State** | [`discover`](#discover), [`contract-create`](#contract-create), [`gate-record`](#gate-record), [`gate-revalidate`](#gate-revalidate), [`activate`](#activate), [`route`](#route), [`preflight`](#preflight), [`advance`](#advance), [`next`](#next), [`record-diagnosis`](#record-diagnosis), [`record-intervention`](#record-intervention), [`record-hypothesis-disposition`](#record-hypothesis-disposition), [`record-decision-criterion`](#record-decision-criterion), [`complete`](#complete), [`clear-state`](#clear-state), [`reconcile-closure`](#reconcile-closure), [`task-create`](#task-create), [`task-list`](#task-list), [`task-show`](#task-show), [`task-lock-status`](#task-lock-status), [`task-scope`](#task-scope) |
|
|
193
|
+
| **Setup & Maintenance** | [`contract-revise`](#contract-revise), [`init`](#init), [`index-setup`](#index-setup), [`index-start`](#index-start), [`index-stop`](#index-stop), [`index-rebuild`](#index-rebuild), [`update`](#update), [`checkpoint-revalidate`](#checkpoint-revalidate), [`task-migrate`](#task-migrate), [`migrate-protocol`](#migrate-protocol), [`task-unlock`](#task-unlock), [`task-recover`](#task-recover), [`task-abandon`](#task-abandon), [`task-repair-contract-bootstrap`](#task-repair-contract-bootstrap), [`task-migrate-contract-bootstrap-repair`](#task-migrate-contract-bootstrap-repair), [`task-repair-legacy-recovery`](#task-repair-legacy-recovery), [`task-resume`](#task-resume) |
|
|
65
194
|
| **Cross-Harness Continuity** | [`continuity`](#continuity), [`record-continuity`](#record-continuity), [`reconcile-continuity`](#reconcile-continuity), [`clear-continuity`](#clear-continuity), [`handoff-create`](#handoff-create), [`handoff-list`](#handoff-list), [`handoff-show`](#handoff-show) |
|
|
66
195
|
| **Durable Actions & Approvals** | [`run-action`](#run-action), [`action-propose`](#action-propose), [`action-record`](#action-record), [`action-show`](#action-show), [`action-reconcile`](#action-reconcile), [`action-verify`](#action-verify), [`action-authorize`](#action-authorize), [`approval-request`](#approval-request), [`approval-resolve`](#approval-resolve) |
|
|
67
196
|
| **Policy & Auditing** | [`policy`](#policy), [`policy-discover`](#policy-discover), [`policy-status`](#policy-status), [`policy-diff`](#policy-diff), [`rule-verify`](#rule-verify), [`baseline`](#baseline), [`bundle`](#bundle) |
|
|
@@ -834,6 +963,107 @@ Updates the managed instruction kit to match the current ForgeLoop package versi
|
|
|
834
963
|
|
|
835
964
|
## 2. Activation & Planning
|
|
836
965
|
|
|
966
|
+
### `discover`
|
|
967
|
+
|
|
968
|
+
Records the canonical initial discovery milestone for a newly created task.
|
|
969
|
+
|
|
970
|
+
- **Purpose**: Transitions a valid post-`task-create` task from `RECEIVED` to the derived `DISCOVERING` phase without creating synthetic work state.
|
|
971
|
+
- **When to use**: When `next` returns `DISCOVER` for a task with no work-state checkpoint.
|
|
972
|
+
- **Mutation**: Appends the task-scoped discovery milestone to the event ledger.
|
|
973
|
+
- **Options**:
|
|
974
|
+
|
|
975
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:discover:options -->
|
|
976
|
+
|
|
977
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
978
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
979
|
+
- `--json`: emit structured discovery output as JSON
|
|
980
|
+
|
|
981
|
+
<!-- END FORGELOOP GENERATED: cli:discover:options -->
|
|
982
|
+
|
|
983
|
+
### `contract-create`
|
|
984
|
+
|
|
985
|
+
Persists a validated contract and materializes the first real lifecycle checkpoint.
|
|
986
|
+
|
|
987
|
+
- **Purpose**: Creates a real contract after discovery and writes `work-state.json` with its actual contract fingerprint.
|
|
988
|
+
- **When to use**: When `next` returns `CREATE_CONTRACT` after discovery.
|
|
989
|
+
- **Mutation**: Writes the task contract, task work state, and append-only lifecycle events transactionally.
|
|
990
|
+
- **Options**:
|
|
991
|
+
|
|
992
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:contract-create:options -->
|
|
993
|
+
|
|
994
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
995
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
996
|
+
- `--contract-file <path>`: validated JSON contract relative to the target
|
|
997
|
+
- `--preset <name>`: bounded contract preset: documentation, bug, feature, or release
|
|
998
|
+
- `--json`: emit structured contract output as JSON
|
|
999
|
+
|
|
1000
|
+
<!-- END FORGELOOP GENERATED: cli:contract-create:options -->
|
|
1001
|
+
|
|
1002
|
+
### `contract-revise`
|
|
1003
|
+
|
|
1004
|
+
Replaces a validated pre-execution contract through an append-only, transaction-witnessed revision.
|
|
1005
|
+
|
|
1006
|
+
- **Purpose**: Canonically revise a contract in `CONTRACT_READY`, `ROUTED`, or `PLANNED` before execution starts.
|
|
1007
|
+
- **When to use**: When the task objective or scope changes and the existing derived route, preflight, gates, or plan must no longer authorize the work.
|
|
1008
|
+
- **Mutation**: Writes the replacement contract, resets derived authorization, appends `CONTRACT_REVISED`, and appends an adjacent `TRANSACTION_COMMITTED(operation=contract-revise)` witness.
|
|
1009
|
+
- **Restrictions**: Requires exactly one of `--preset` or `--contract-file`; a `PLANNED` task rewinds to `ROUTED`, and old route/preflight/plan evidence must be regenerated.
|
|
1010
|
+
- **Options**:
|
|
1011
|
+
|
|
1012
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:contract-revise:options -->
|
|
1013
|
+
|
|
1014
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
1015
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
1016
|
+
- `--contract-file <path>`: validated replacement JSON contract relative to the target
|
|
1017
|
+
- `--preset <name>`: bounded replacement contract preset: documentation, bug, feature, or release
|
|
1018
|
+
- `--json`: emit structured contract revision output as JSON
|
|
1019
|
+
|
|
1020
|
+
<!-- END FORGELOOP GENERATED: cli:contract-revise:options -->
|
|
1021
|
+
|
|
1022
|
+
### `gate-record`
|
|
1023
|
+
|
|
1024
|
+
Records a gate satisfaction or rejection decision for a task's preflight gate lifecycle.
|
|
1025
|
+
|
|
1026
|
+
- **Purpose**: Records structured evidence that a named gate has been satisfied, rejected, or deferred, with optional artifact and decision evidence.
|
|
1027
|
+
- **When to use**: During the preflight gate lifecycle to progress gate status from pending to resolved.
|
|
1028
|
+
- **Mutation**: Writes a gate artifact under the task namespace and appends a `GATE_SATISFIED` or `GATE_REJECTED` protocol event. A `GATE_SATISFIED` event after `CONTRACT_REVISED` must be immediately followed by the canonical `TRANSACTION_COMMITTED(operation=gate-record)` witness; pre-revision historical gate events remain backward compatible. Repeated satisfied recording is event-idempotent within the current contract epoch, even if the gate artifact is refreshed.
|
|
1029
|
+
- **Options**:
|
|
1030
|
+
|
|
1031
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:gate-record:options -->
|
|
1032
|
+
|
|
1033
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
1034
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
1035
|
+
- `--gate <name>`: required gate name
|
|
1036
|
+
- `--status <status>`: satisfied, unverified, or blocked
|
|
1037
|
+
- `--artifact <path>`: project-relative evidence artifact (repeatable)
|
|
1038
|
+
- `--decision <text>`: caller-recorded gate decision (repeatable)
|
|
1039
|
+
- `--unknown <text>`: known unresolved item (repeatable)
|
|
1040
|
+
- `--assumption <text>`: approved local assumption (repeatable)
|
|
1041
|
+
- `--evidence-file <path>`: bounded local descriptive evidence JSON
|
|
1042
|
+
- `--json`: emit structured gate output as JSON
|
|
1043
|
+
|
|
1044
|
+
<!-- END FORGELOOP GENERATED: cli:gate-record:options -->
|
|
1045
|
+
|
|
1046
|
+
### `gate-revalidate`
|
|
1047
|
+
|
|
1048
|
+
Refreshes a satisfied gate whose evidence artifacts changed after execution
|
|
1049
|
+
started, while preserving the original approval and append-only history.
|
|
1050
|
+
|
|
1051
|
+
- **Purpose**: Revalidate stale gate artifacts only through current task identity and active write claims.
|
|
1052
|
+
- **When to use**: When `next` returns `REVALIDATE_GATES` for a post-execution task.
|
|
1053
|
+
- **Mutation**: Refreshes the task-scoped gate artifact and appends `GATE_REVALIDATED` with an adjacent `TRANSACTION_COMMITTED(operation=gate-revalidate)` witness.
|
|
1054
|
+
- **Restrictions**: Does not bypass `gate-record`'s `E_PHASE_FREEZE`; requires `--acknowledge-stale`, a coherent current route, and changed artifacts covered by active claims.
|
|
1055
|
+
- **Options**:
|
|
1056
|
+
|
|
1057
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:gate-revalidate:options -->
|
|
1058
|
+
|
|
1059
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
1060
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
1061
|
+
- `--gate <name>`: satisfied gate to refresh after safe artifact drift
|
|
1062
|
+
- `--acknowledge-stale`: explicitly acknowledge refresh of stale gate artifacts inside active task claims
|
|
1063
|
+
- `--json`: emit structured gate revalidation output as JSON
|
|
1064
|
+
|
|
1065
|
+
<!-- END FORGELOOP GENERATED: cli:gate-revalidate:options -->
|
|
1066
|
+
|
|
837
1067
|
### `route`
|
|
838
1068
|
|
|
839
1069
|
Calculates and persists deterministic engineering guide routing.
|
|
@@ -1967,6 +2197,31 @@ Clears canonical work-state checkpoint for the current task.
|
|
|
1967
2197
|
forgeloop clear-state
|
|
1968
2198
|
```
|
|
1969
2199
|
|
|
2200
|
+
### `checkpoint-revalidate`
|
|
2201
|
+
|
|
2202
|
+
Revalidates a safe pre-execution `ROUTED` checkpoint after repository-only
|
|
2203
|
+
drift.
|
|
2204
|
+
|
|
2205
|
+
- **Purpose**: Rebinds the checkpoint to ForgeLoop's current repository fingerprint while preserving lifecycle, contract, route, guide, gate, check, and evidence identity.
|
|
2206
|
+
- **When to use**: When `next` returns `REVALIDATE_CHECKPOINT` with only `E_REPOSITORY_CHANGED` / `E_STATE_REVALIDATION_REQUIRED`.
|
|
2207
|
+
- **Mutation**: Advances the state revision and appends a transaction-bound `CHECKPOINT_REVALIDATED` event.
|
|
2208
|
+
- **Safety Note**: It does not accept caller-supplied repository identity, revise contracts, reroute, fabricate evidence, or operate after execution starts. Unsupported drift fails closed with `E_CHECKPOINT_REVALIDATION_UNSAFE`.
|
|
2209
|
+
- **Options**:
|
|
2210
|
+
|
|
2211
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:checkpoint-revalidate:options -->
|
|
2212
|
+
|
|
2213
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
2214
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
2215
|
+
- `--json`: emit structured revalidation output as JSON
|
|
2216
|
+
|
|
2217
|
+
<!-- END FORGELOOP GENERATED: cli:checkpoint-revalidate:options -->
|
|
2218
|
+
|
|
2219
|
+
- **Example**:
|
|
2220
|
+
|
|
2221
|
+
```bash
|
|
2222
|
+
forgeloop checkpoint-revalidate --task <id> --json
|
|
2223
|
+
```
|
|
2224
|
+
|
|
1970
2225
|
---
|
|
1971
2226
|
|
|
1972
2227
|
### `reconcile-closure`
|
|
@@ -1974,9 +2229,9 @@ Clears canonical work-state checkpoint for the current task.
|
|
|
1974
2229
|
Reconciles the checkpoint of an EXECUTING, VERIFYING, or REVIEWING task whose objective is already satisfied in the current repository.
|
|
1975
2230
|
|
|
1976
2231
|
- **Purpose**: Refresh the work-state repository fingerprint of a stale EXECUTING, VERIFYING, or REVIEWING task after repository movement, using executed contract-bound evidence that the objective is present, so the canonical completion pipeline can close it.
|
|
1977
|
-
- **When to use**: When a task is stuck in EXECUTING, VERIFYING, or REVIEWING with `E_REPOSITORY_CHANGED` / `E_STATE_REVALIDATION_REQUIRED` and its objective was already satisfied by other changes in the current repository. A REVIEWING task
|
|
2232
|
+
- **When to use**: When a task is stuck in EXECUTING, VERIFYING, or REVIEWING with `E_REPOSITORY_CHANGED` / `E_STATE_REVALIDATION_REQUIRED` and its objective was already satisfied by other changes in the current repository. A REVIEWING task may use an existing authorized completion-recovery snapshot, or the narrow bootstrap path when the only drift is repository movement and no completion rejection has been persisted.
|
|
1978
2233
|
- **Mutation**: Appends a `CHECKPOINT_RECONCILED` ledger event (previous/current repository fingerprints plus evidence) and refreshes the work-state repository fingerprint. The phase stays unchanged until the canonical pipeline advances it; claims release only through canonical `COMPLETE`.
|
|
1979
|
-
- **Safety Note**: Refuses other phases, fresh checkpoints, contract or artifact drift, invalid ledgers, unknown requirements, and failing evidence.
|
|
2234
|
+
- **Safety Note**: Refuses other phases, fresh checkpoints, contract or required-artifact drift, invalid ledgers, invalid claim ownership, unauthorized persisted completion rejection, unknown requirements, and failing evidence. It never appends completion events or releases claims.
|
|
1980
2235
|
- **Options**:
|
|
1981
2236
|
|
|
1982
2237
|
<!-- BEGIN FORGELOOP GENERATED: cli:reconcile-closure:options -->
|
|
@@ -2251,6 +2506,111 @@ Fake, missing, corrupt, or mismatched recovery state is
|
|
|
2251
2506
|
`E_TASK_CLAIM_OWNERSHIP_INCONSISTENT`/`E_TASK_RECOVERY_INCONSISTENT`; historical
|
|
2252
2507
|
claims remain reserved.
|
|
2253
2508
|
|
|
2509
|
+
### `task-abandon`
|
|
2510
|
+
|
|
2511
|
+
Explicitly abandons an active non-terminal task without fabricating completion.
|
|
2512
|
+
|
|
2513
|
+
- **Purpose**: Releases validated write claims for a deliberately abandoned task while preserving its current lifecycle phase, append-only history, and durable recovery boundary. This is the canonical escape from an active-task claim deadlock; it is not a completion or publication operation.
|
|
2514
|
+
- **When to use**: Only when the exact task ID is known and the caller has deliberately acknowledged abandonment. Use `task-recover` only for tasks already classified `STALE` or `ABANDONED`.
|
|
2515
|
+
- **Mutation**: Appends `TASK_ABANDONED` and its `TRANSACTION_COMMITTED(operation=task-abandon)` witness, then writes `recovery.json` with classification `ABANDONED` and authority `CALLER_ACKNOWLEDGED`. Work state remains unchanged and claims resolve to `RELEASED_BY_RECOVERY`.
|
|
2516
|
+
- **Options**:
|
|
2517
|
+
|
|
2518
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:task-abandon:options -->
|
|
2519
|
+
|
|
2520
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
2521
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
2522
|
+
- `--acknowledge-abandonment`: explicitly acknowledge abandonment of an active task (required; not completion authority)
|
|
2523
|
+
- `--json`: emit structured abandonment output as JSON
|
|
2524
|
+
|
|
2525
|
+
<!-- END FORGELOOP GENERATED: cli:task-abandon:options -->
|
|
2526
|
+
|
|
2527
|
+
- **Example**:
|
|
2528
|
+
|
|
2529
|
+
```bash
|
|
2530
|
+
forgeloop task-abandon --task task-001 --acknowledge-abandonment --json
|
|
2531
|
+
```
|
|
2532
|
+
|
|
2533
|
+
The command requires an explicit task ID and acknowledgement. It refuses
|
|
2534
|
+
`COMPLETE`, already recovered, inconsistent, or non-active ownership states
|
|
2535
|
+
and never writes `COMPLETION_VALIDATED`. Use `task-resume` to reacquire claims
|
|
2536
|
+
through the normal lifecycle when work should continue.
|
|
2537
|
+
|
|
2538
|
+
### `task-repair-contract-bootstrap`
|
|
2539
|
+
|
|
2540
|
+
Repairs only the exact historical duplicate contract bootstrap defect.
|
|
2541
|
+
|
|
2542
|
+
- **Purpose**: Recognizes the narrow append-only signature of a duplicate `CONTRACT_VALIDATED` followed by a duplicate `contract-create` commit, verifies the current contract and route/state bindings, and reconstructs the earliest proven checkpoint without rewriting existing events.
|
|
2543
|
+
- **Mutation**: Under project/task serialization, writes the reconciled work-state and appends `CONTRACT_BOOTSTRAP_REPAIR_RECORDED` plus `TRANSACTION_COMMITTED` in one transaction.
|
|
2544
|
+
- **Options**:
|
|
2545
|
+
|
|
2546
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:task-repair-contract-bootstrap:options -->
|
|
2547
|
+
|
|
2548
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
2549
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
2550
|
+
- `--acknowledge-repair`: explicit caller acknowledgement of the exact append-only repair (required)
|
|
2551
|
+
- `--json`: emit structured repair output as JSON
|
|
2552
|
+
|
|
2553
|
+
<!-- END FORGELOOP GENERATED: cli:task-repair-contract-bootstrap:options -->
|
|
2554
|
+
|
|
2555
|
+
- **Example**:
|
|
2556
|
+
|
|
2557
|
+
```bash
|
|
2558
|
+
forgeloop task-repair-contract-bootstrap --task task-001 --acknowledge-repair --json
|
|
2559
|
+
```
|
|
2560
|
+
|
|
2561
|
+
The command requires fresh caller acknowledgement. The marker records the
|
|
2562
|
+
repair-time checkpoint as an immutable anchor, including `reconstructedPhase`,
|
|
2563
|
+
`reconstructedStateFingerprint`, `reconstructedStateRevision`, and the
|
|
2564
|
+
repair-time `routeFingerprint`; it does not freeze the task at that checkpoint.
|
|
2565
|
+
The marker must be immediately followed by its
|
|
2566
|
+
`TRANSACTION_COMMITTED(operation=task-repair-contract-bootstrap)` witness.
|
|
2567
|
+
After the anchor revision, normal canonical lifecycle evolution is allowed, but
|
|
2568
|
+
state rollback, missing state, invalid current contract/route identity, or
|
|
2569
|
+
ledger/state incoherence fails closed with `E_CONTRACT_BOOTSTRAP_REPAIR_INVALID`.
|
|
2570
|
+
If a later route fingerprint differs from the marker, the current route and
|
|
2571
|
+
state must be canonically coherent and the ledger must contain a ForgeLoop-
|
|
2572
|
+
generated `ROUTE_REBOUND` witness immediately followed by the matching
|
|
2573
|
+
`TRANSACTION_COMMITTED(operation=route)`. The witness binds the current route
|
|
2574
|
+
fingerprint, the repair-time contract fingerprint, and the previous route
|
|
2575
|
+
fingerprint; a historical route transaction, or route and state fields changed
|
|
2576
|
+
together without this identity witness, is not authorization. A contract-only
|
|
2577
|
+
repair may establish its first route through canonical `runRoute`, recording a
|
|
2578
|
+
null previous route fingerprint. A proven `ROUTE_VALIDATED` milestone also
|
|
2579
|
+
requires a present, valid route artifact bound to the current contract, while a
|
|
2580
|
+
history without `ROUTE_VALIDATED` may repair to `CONTRACT_READY` without a
|
|
2581
|
+
route artifact. After repair, entering `DESIGNING` additionally requires the
|
|
2582
|
+
canonical `DESIGN_GATE_STARTED` event.
|
|
2583
|
+
|
|
2584
|
+
### `task-migrate-contract-bootstrap-repair`
|
|
2585
|
+
|
|
2586
|
+
Migrates the exact legacy contract bootstrap repair marker produced before
|
|
2587
|
+
`reconstructedStateRevision` was required.
|
|
2588
|
+
|
|
2589
|
+
- **Purpose**: Proves the legacy marker, its immediate repair transaction, the current contract, the reconstructed work-state, and (for `ROUTED`) the persisted route. Strict validation remains fail-closed until the migration witness is appended.
|
|
2590
|
+
- **Mutation**: Appends `CONTRACT_BOOTSTRAP_REPAIR_MIGRATION_RECORDED` plus `TRANSACTION_COMMITTED(operation=task-migrate-contract-bootstrap-repair)`; it never rewrites the legacy marker or state/contract/route artifacts.
|
|
2591
|
+
- **Options**:
|
|
2592
|
+
|
|
2593
|
+
<!-- BEGIN FORGELOOP GENERATED: cli:task-migrate-contract-bootstrap-repair:options -->
|
|
2594
|
+
|
|
2595
|
+
- `--path <directory>`: target project directory (default: current directory)
|
|
2596
|
+
- `--task <id>`: task ID to operate on (when omitted, resolved from context or single active task)
|
|
2597
|
+
- `--acknowledge-migration`: fresh explicit caller acknowledgement of the exact legacy marker migration (required)
|
|
2598
|
+
- `--json`: emit structured migration output as JSON
|
|
2599
|
+
|
|
2600
|
+
<!-- END FORGELOOP GENERATED: cli:task-migrate-contract-bootstrap-repair:options -->
|
|
2601
|
+
|
|
2602
|
+
- **Example**:
|
|
2603
|
+
|
|
2604
|
+
```bash
|
|
2605
|
+
forgeloop task-migrate-contract-bootstrap-repair --task task-001 --acknowledge-migration --json
|
|
2606
|
+
```
|
|
2607
|
+
|
|
2608
|
+
Only the exact legacy schema and immediate repair boundary are eligible. Any
|
|
2609
|
+
near-miss, later lifecycle activity, state/route drift, live lock, unknown lock,
|
|
2610
|
+
or corrupt lock fails closed with
|
|
2611
|
+
`E_CONTRACT_BOOTSTRAP_REPAIR_MIGRATION_INVALID`/`E_TASK_LOCKED`. The operation
|
|
2612
|
+
is idempotent only when the complete migrated relationship remains valid.
|
|
2613
|
+
|
|
2254
2614
|
### `task-resume`
|
|
2255
2615
|
|
|
2256
2616
|
Reacquires a recovered task's write claims and restores ordinary mutation authority.
|
package/docs/CODE_ATTESTATION.md
CHANGED
|
@@ -9,11 +9,11 @@ bug-free or secure.
|
|
|
9
9
|
|
|
10
10
|
| Level | Meaning |
|
|
11
11
|
| --- | --- |
|
|
12
|
-
| `PROCESSED` |
|
|
12
|
+
| `PROCESSED` | Minimum reported level; verification is not complete. `MISSING`, `DISABLED`, and `INVALID` remain separate status results. |
|
|
13
13
|
| `VERIFIED` | Completion, evidence bindings, manifest, and current content validate. |
|
|
14
14
|
| `ATTESTED` | `VERIFIED` plus a cryptographically valid signature and trusted signer policy. |
|
|
15
15
|
|
|
16
|
-
`PROCESSED` is
|
|
16
|
+
`PROCESSED` is a lower-bound reported level, not a trust claim. A manifest or
|
|
17
17
|
statement that merely exists never becomes `ATTESTED`; the signature must be
|
|
18
18
|
verified against the configured signer identity, issuer, and trust policy.
|
|
19
19
|
Attestation binds bounded source/evidence relationships. It does not prove
|
|
@@ -54,11 +54,30 @@ Documentation-impact questions for integration/MCP changes:
|
|
|
54
54
|
- Did an adapter error code change?
|
|
55
55
|
- Did an integration limit or resource list change?
|
|
56
56
|
|
|
57
|
-
Anti-drift invariant: every
|
|
58
|
-
`
|
|
59
|
-
|
|
60
|
-
|
|
61
|
-
|
|
57
|
+
Anti-drift invariant: every packaged Markdown or harness-adapter document in
|
|
58
|
+
`package.json.files` is registered in `documentation-manifest.json`, and every
|
|
59
|
+
`packaged: true` manifest entry is mechanically checked against the core npm
|
|
60
|
+
tarball contents (`tests/package.test.js`). The review matrix records generated
|
|
61
|
+
or handwritten origin, current or historical status, package inclusion, action,
|
|
62
|
+
and canonical sources for every registered document.
|
|
63
|
+
|
|
64
|
+
### Documentation retention and repository hygiene
|
|
65
|
+
|
|
66
|
+
Use one canonical owner for each maintained concept. Generated documentation
|
|
67
|
+
must name its source, generator, freshness check, and reason for being tracked.
|
|
68
|
+
Historical audits, release evidence, and completed validation records belong
|
|
69
|
+
under `docs/history/`; current operating instructions must not leave those
|
|
70
|
+
records loose in the repository root. One-off plans and validation reports are
|
|
71
|
+
either consolidated into a canonical document and deleted or moved to the
|
|
72
|
+
history index with their current owner links.
|
|
73
|
+
|
|
74
|
+
Repository and npm decisions are separate. GitHub may retain reproducible
|
|
75
|
+
benchmarks, PoC evidence, source-bound diagram artifacts, vendored renderer
|
|
76
|
+
sources, and historical records that consumers do not need. The npm package
|
|
77
|
+
ships only intentional runtime, integration, legal, harness, and canonical user
|
|
78
|
+
documentation surfaces. `npm run repository:hygiene` enforces explicit root,
|
|
79
|
+
tracked-state, visual-ownership, benchmark-run-set, and scratch-output policy;
|
|
80
|
+
it does not guess whether an arbitrary document is useful.
|
|
62
81
|
and `WORK_TRANSITIONS`; do not maintain independent hand-written transition
|
|
63
82
|
enums when a generated or mechanically validated representation is available.
|
|
64
83
|
|
|
@@ -195,7 +214,7 @@ migration, or security-sensitive require `npm run docs:check` before merge.
|
|
|
195
214
|
2. **Pinned local renderer**: Generation uses only the vendored Archify v2.15.0 source at the reviewed commit recorded in `docs/diagrams/manifest.json` and `vendor/archify/v2.15.0/PIN.json`.
|
|
196
215
|
3. **Animated committed outputs**: Every active source uses `meta.animation: "trace"`. Each interactive HTML is the primary animated explorer, and each self-contained SVG fallback carries trace-capable edge/node animation while remaining usable in repository previews. Deterministic receipts are committed under `docs/assets/diagrams/`.
|
|
197
216
|
4. **GitHub-safe SVG**: The SVG must not embed `<script>` or `<foreignObject>`, must expose accessible title/description metadata, and must remain visible through standard Markdown image syntax.
|
|
198
|
-
5. **Fingerprint and review verification**: The generated SVG embeds a `data-forgeloop-source-sha256` attribute, the outputs expose trace markers, and the receipt binds the source, HTML, and SVG hashes. The
|
|
217
|
+
5. **Fingerprint and review verification**: The generated SVG embeds a `data-forgeloop-source-sha256` attribute, the outputs expose trace markers, and the receipt binds the source, HTML, and SVG hashes. The review at `docs/diagrams/reviews/` binds the current source and SVG hashes as a review assertion and is never generated or overwritten; the checker does not independently authenticate the reviewer. Run `npm run docs:diagrams:check` before review.
|
|
199
218
|
6. **Scoped wrapper**: The ForgeLoop Archify wrapper is intentionally documentation-scoped. It reads canonical inputs only from `docs/diagrams/` and permits deliver outputs only under `docs/assets/diagrams/`.
|
|
200
219
|
|
|
201
220
|
ForgeLoop governs five documentation-diagram categories: workflow,
|
|
@@ -213,14 +232,16 @@ README hero assets are branding/conceptual architecture illustrations. They are
|
|
|
213
232
|
not the canonical protocol diagram. The typed Archify workflow under
|
|
214
233
|
`docs/diagrams/` remains the canonical lifecycle architecture source, with
|
|
215
234
|
generated outputs under `docs/assets/diagrams/`; the CLI-only persistent search
|
|
216
|
-
transport is explained by `docs/PERSISTENT_SEARCH_TRANSPORT.md`.
|
|
235
|
+
transport is explained by `docs/PERSISTENT_SEARCH_TRANSPORT.md`. The current
|
|
236
|
+
repository-only assets are `docs/assets/forgeloop-architecture.svg` and
|
|
237
|
+
`docs/assets/forgeloop-lifecycle-animated.svg`; neither is a package file.
|
|
217
238
|
|
|
218
239
|
The README hero is intentionally GitHub-repository-only:
|
|
219
240
|
|
|
220
|
-
- `README.md`
|
|
221
|
-
|
|
222
|
-
|
|
223
|
-
|
|
241
|
+
- `README.md` references `docs/assets/forgeloop-architecture.svg` for the hero
|
|
242
|
+
and `docs/assets/forgeloop-lifecycle-animated.svg` for the looping overview;
|
|
243
|
+
both are repository-only and excluded from the npm package.
|
|
244
|
+
- `tests/package.test.js` asserts those exclusions so they cannot be silently
|
|
224
245
|
re-included.
|
|
225
246
|
- The packaged README is therefore not self-contained for that relative hero
|
|
226
247
|
path; do not claim otherwise.
|
|
@@ -0,0 +1,31 @@
|
|
|
1
|
+
# Jev benchmark and calibration
|
|
2
|
+
|
|
3
|
+
`npm run benchmark:jev` emits an offline deterministic baseline for model-route
|
|
4
|
+
scenarios. It records the pinned engine/model, safety-floor projection, fallback
|
|
5
|
+
status, Jev/cache/latency fields, context and tool dimensions, and unknown host
|
|
6
|
+
telemetry. Zero counts mean that no provider request was made; `null` and
|
|
7
|
+
`NOT_MEASURED` remain explicit when the host did not observe a value. It does
|
|
8
|
+
not fabricate provider usage and does not authorize lifecycle, execution,
|
|
9
|
+
pruning, or completion.
|
|
10
|
+
|
|
11
|
+
Provider-backed calibration requires a live TypeSafe organization with credits;
|
|
12
|
+
an unavailable provider is reported as unavailable rather than treated as a
|
|
13
|
+
successful benchmark.
|
|
14
|
+
|
|
15
|
+
`npm run benchmark:jev:live` runs bounded intake, route, and context requests
|
|
16
|
+
through the pinned Jev provider and reports provider usage/latency only when the
|
|
17
|
+
provider supplies it. It requires `TYPESAFE_API_KEY`, never prints that key,
|
|
18
|
+
and is intentionally separate from the offline benchmark.
|
|
19
|
+
|
|
20
|
+
An optional maintainer release check is available through
|
|
21
|
+
`npm run jev:smoke` and `npm run benchmark:jev:live`. The current repository does
|
|
22
|
+
not enforce either live command in CI or the release workflow, and the CLI does
|
|
23
|
+
not bind either command to an exact candidate commit. Offline output, a missing
|
|
24
|
+
credential, or a provider-unavailable result is never a substitute for live
|
|
25
|
+
interoperability evidence.
|
|
26
|
+
|
|
27
|
+
Dependency audit attribution for the current base and PR head is unchanged:
|
|
28
|
+
one high-severity `js-yaml` advisory (`GHSA-2883-xcg3-v3hh`, CVSS 7.5, CWE-400
|
|
29
|
+
and CWE-407) arrives transitively through `eslint` → `@eslint/eslintrc` →
|
|
30
|
+
`js-yaml` 4.3.1. It is not introduced by the Jev dependency; the audit reports
|
|
31
|
+
an available upgrade outside this correction's runtime dependency policy.
|
|
@@ -0,0 +1,37 @@
|
|
|
1
|
+
# Model routing
|
|
2
|
+
|
|
3
|
+
ForgeLoop exposes model routing as a bounded semantic-decision projection. It
|
|
4
|
+
has `SEMANTIC_DECISION` authority only within the declared decision contract;
|
|
5
|
+
it has no lifecycle, completion, evidence, ownership, installation, command, or
|
|
6
|
+
publication authority. The deterministic policy establishes the safety floor:
|
|
7
|
+
|
|
8
|
+
- `NONE` when no generation is required;
|
|
9
|
+
- `STANDARD` for ordinary executable generation;
|
|
10
|
+
- `PRIMARY` for security, architecture, migration, ambiguity, and other
|
|
11
|
+
high-risk signals.
|
|
12
|
+
|
|
13
|
+
The deterministic floor currently emits `NONE`, `STANDARD`, or `PRIMARY`. The
|
|
14
|
+
public vocabulary also includes `FAST`, which is available only as an advisory
|
|
15
|
+
escalation and cannot lower the deterministic floor.
|
|
16
|
+
|
|
17
|
+
The pinned Jev model (`jev-1.13.0`) may recommend an escalation or request an
|
|
18
|
+
escalation when confidence is low. It can never lower the deterministic floor,
|
|
19
|
+
select a vendor-specific model, execute a command, change lifecycle state,
|
|
20
|
+
authorize ownership, or weaken verification requirements. ForgeLoop remains the
|
|
21
|
+
authority for all lifecycle, evidence, safety, and completion decisions.
|
|
22
|
+
|
|
23
|
+
For route execution, the same boundary applies to guide relevance: Jev can
|
|
24
|
+
reorder or remove a selected non-mandatory guide only with sufficient
|
|
25
|
+
confidence. Mandatory safety protection is derived from the canonical
|
|
26
|
+
deterministic route reasons the router already produced (auth surface and the
|
|
27
|
+
trust-boundary risks untrusted-input, personal-data, secrets, external-service,
|
|
28
|
+
publication), so a `security` guide selected by `external-service` risk cannot be
|
|
29
|
+
removed. Mandatory safety guides are retained and low-confidence removal
|
|
30
|
+
recommendations are retained rather than treated as authority. The resulting
|
|
31
|
+
guide set and profile are persisted in the route artifact, so the semantic
|
|
32
|
+
recommendation materially affects routing without becoming lifecycle authority.
|
|
33
|
+
|
|
34
|
+
The live Jev provider is not an authority substitute. Semantic-required
|
|
35
|
+
model-routing operations consume a fresh persisted `MODEL_ROUTE` decision and
|
|
36
|
+
fail closed when it is unavailable or stale; offline inspection may still
|
|
37
|
+
project deterministic policy without making a network request.
|