bmad-speckit-sdd-flow 1.0.2 → 1.0.3
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/README.md +16 -14
- package/README.zh-CN.md +16 -14
- package/_bmad/_config/governance-remediation.yaml +26 -0
- package/_bmad/claude/agents/auditors/auditor-bugfix.md +6 -0
- package/_bmad/claude/agents/auditors/auditor-spec.md +8 -2
- package/_bmad/claude/agents/auditors/auditor-tasks-doc.md +6 -0
- package/_bmad/claude/agents/bmad-epic-audit.md +4 -1
- package/_bmad/claude/agents/bmad-master.md +39 -24
- package/_bmad/claude/agents/bmad-story-audit.md +5 -1
- package/_bmad/claude/agents/bmad-story-create.md +3 -1
- package/_bmad/claude/agents/layers/bmad-bug-agent.md +7 -1
- package/_bmad/claude/agents/layers/bmad-code-reviewer-lifecycle.md +4 -0
- package/_bmad/claude/agents/layers/bmad-layer4-speckit-gaps.md +4 -1
- package/_bmad/claude/agents/layers/bmad-layer4-speckit-implement.md +4 -1
- package/_bmad/claude/agents/layers/bmad-layer4-speckit-plan.md +4 -1
- package/_bmad/claude/agents/layers/bmad-layer4-speckit-specify.md +9 -1
- package/_bmad/claude/agents/layers/bmad-layer4-speckit-tasks.md +4 -1
- package/_bmad/claude/agents/layers/bmad-standalone-tasks.md +7 -1
- package/_bmad/claude/hooks/post-tool-use.cjs +8 -43
- package/_bmad/claude/hooks/stop.cjs +8 -28
- package/_bmad/claude/protocols/handoff-schema.md +39 -0
- package/_bmad/claude/skills/bmad-bug-assistant/SKILL.en.md +24 -0
- package/_bmad/claude/skills/bmad-bug-assistant/SKILL.md +36 -1
- package/_bmad/claude/skills/bmad-bug-assistant/SKILL.zh.md +18 -0
- package/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.en.md +23 -1
- package/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.md +195 -169
- package/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.zh.md +18 -0
- package/_bmad/claude/skills/bmad-rca-helper/SKILL.en.md +18 -0
- package/_bmad/claude/skills/bmad-rca-helper/SKILL.md +161 -142
- package/_bmad/claude/skills/bmad-rca-helper/SKILL.zh.md +16 -0
- package/_bmad/claude/skills/bmad-standalone-tasks/SKILL.en.md +44 -0
- package/_bmad/claude/skills/bmad-standalone-tasks/SKILL.md +66 -3
- package/_bmad/claude/skills/bmad-standalone-tasks/SKILL.zh.md +38 -0
- package/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.en.md +33 -2
- package/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.md +165 -118
- package/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.zh.md +31 -1
- package/_bmad/claude/skills/bmad-story-assistant/SKILL.en.md +40 -0
- package/_bmad/claude/skills/bmad-story-assistant/SKILL.md +52 -1
- package/_bmad/claude/skills/bmad-story-assistant/SKILL.zh.md +34 -0
- package/_bmad/claude/skills/speckit-workflow/SKILL.en.md +35 -0
- package/_bmad/claude/skills/speckit-workflow/SKILL.md +27 -1
- package/_bmad/claude/skills/speckit-workflow/SKILL.zh.md +27 -1
- package/_bmad/cursor/hooks/post-tool-use.cjs +7 -42
- package/_bmad/cursor/skills/bmad-bug-assistant/SKILL.en.md +18 -0
- package/_bmad/cursor/skills/bmad-bug-assistant/SKILL.md +30 -1
- package/_bmad/cursor/skills/bmad-bug-assistant/SKILL.zh.md +16 -1
- package/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.en.md +15 -1
- package/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.md +57 -38
- package/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.zh.md +14 -0
- package/_bmad/cursor/skills/bmad-rca-helper/SKILL.en.md +14 -0
- package/_bmad/cursor/skills/bmad-rca-helper/SKILL.md +77 -62
- package/_bmad/cursor/skills/bmad-rca-helper/SKILL.zh.md +14 -0
- package/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.en.md +18 -0
- package/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.md +39 -3
- package/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.zh.md +16 -0
- package/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.en.md +14 -0
- package/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.md +54 -38
- package/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.zh.md +14 -0
- package/_bmad/cursor/skills/bmad-story-assistant/SKILL.en.md +18 -0
- package/_bmad/cursor/skills/bmad-story-assistant/SKILL.md +37 -8
- package/_bmad/cursor/skills/bmad-story-assistant/SKILL.zh.md +16 -1
- package/_bmad/cursor/skills/speckit-workflow/SKILL.en.md +29 -0
- package/_bmad/cursor/skills/speckit-workflow/SKILL.md +30 -2
- package/_bmad/cursor/skills/speckit-workflow/SKILL.zh.md +30 -2
- package/_bmad/runtime/hooks/governance-packet-hard-closeout.cjs +172 -49
- package/_bmad/runtime/hooks/party-mode-read-current-session.cjs +1 -2
- package/_bmad/runtime/hooks/pre-continue-check.cjs +225 -1
- package/_bmad/runtime/hooks/runtime-policy-inject-core.cjs +536 -7
- package/_bmad/skills/consumer-governance-validation/SKILL.md +30 -69
- package/_bmad/skills/consumer-governance-validation/references/consumer-governance-validation.md +34 -258
- package/docs/assets/readme-architecture-overview.final.svg +54 -8
- package/docs/assets/readme-slogan.final.svg +10 -51
- package/node_modules/@bmad-speckit/ralph-method/package.json +1 -1
- package/node_modules/@bmad-speckit/runtime-context/package.json +1 -1
- package/node_modules/@bmad-speckit/runtime-emit/dist/auditor-post-actions.cjs +6 -0
- package/node_modules/@bmad-speckit/runtime-emit/dist/emit-runtime-policy.cjs +2625 -408
- package/node_modules/@bmad-speckit/runtime-emit/dist/resolve-for-session.cjs +3005 -909
- package/node_modules/@bmad-speckit/runtime-emit/dist/run-auditor-host.cjs +259 -40
- package/node_modules/@bmad-speckit/runtime-emit/dist/update-runtime-audit-index.cjs +6 -0
- package/node_modules/@bmad-speckit/runtime-emit/package.json +2 -7
- package/node_modules/@bmad-speckit/schema/package.json +1 -1
- package/node_modules/@bmad-speckit/scoring/dist/dashboard/mcp-server.js +1 -1
- package/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.d.ts +2 -1
- package/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.d.ts.map +1 -1
- package/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.js +17 -2
- package/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.js.map +1 -1
- package/node_modules/@bmad-speckit/scoring/package.json +1 -1
- package/node_modules/bmad-speckit/_bmad/_config/governance-remediation.yaml +26 -0
- package/node_modules/bmad-speckit/_bmad/claude/agents/auditors/auditor-bugfix.md +6 -0
- package/node_modules/bmad-speckit/_bmad/claude/agents/auditors/auditor-spec.md +8 -2
- package/node_modules/bmad-speckit/_bmad/claude/agents/auditors/auditor-tasks-doc.md +6 -0
- package/node_modules/bmad-speckit/_bmad/claude/agents/bmad-epic-audit.md +4 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/bmad-master.md +39 -24
- package/node_modules/bmad-speckit/_bmad/claude/agents/bmad-story-audit.md +5 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/bmad-story-create.md +3 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-bug-agent.md +7 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-code-reviewer-lifecycle.md +4 -0
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-layer4-speckit-gaps.md +4 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-layer4-speckit-implement.md +4 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-layer4-speckit-plan.md +4 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-layer4-speckit-specify.md +9 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-layer4-speckit-tasks.md +4 -1
- package/node_modules/bmad-speckit/_bmad/claude/agents/layers/bmad-standalone-tasks.md +7 -1
- package/node_modules/bmad-speckit/_bmad/claude/hooks/post-tool-use.cjs +8 -43
- package/node_modules/bmad-speckit/_bmad/claude/hooks/stop.cjs +8 -28
- package/node_modules/bmad-speckit/_bmad/claude/protocols/handoff-schema.md +39 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-bug-assistant/SKILL.en.md +24 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-bug-assistant/SKILL.md +36 -1
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-bug-assistant/SKILL.zh.md +18 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.en.md +23 -1
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.md +195 -169
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-code-reviewer-lifecycle/SKILL.zh.md +18 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-rca-helper/SKILL.en.md +18 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-rca-helper/SKILL.md +161 -142
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-rca-helper/SKILL.zh.md +16 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks/SKILL.en.md +44 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks/SKILL.md +66 -3
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks/SKILL.zh.md +38 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.en.md +33 -2
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.md +165 -118
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-standalone-tasks-doc-review/SKILL.zh.md +31 -1
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-story-assistant/SKILL.en.md +40 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-story-assistant/SKILL.md +52 -1
- package/node_modules/bmad-speckit/_bmad/claude/skills/bmad-story-assistant/SKILL.zh.md +34 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/speckit-workflow/SKILL.en.md +35 -0
- package/node_modules/bmad-speckit/_bmad/claude/skills/speckit-workflow/SKILL.md +27 -1
- package/node_modules/bmad-speckit/_bmad/claude/skills/speckit-workflow/SKILL.zh.md +27 -1
- package/node_modules/bmad-speckit/_bmad/cursor/hooks/post-tool-use.cjs +7 -42
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-bug-assistant/SKILL.en.md +18 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-bug-assistant/SKILL.md +30 -1
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-bug-assistant/SKILL.zh.md +16 -1
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.en.md +15 -1
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.md +57 -38
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-code-reviewer-lifecycle/SKILL.zh.md +14 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-rca-helper/SKILL.en.md +14 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-rca-helper/SKILL.md +77 -62
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-rca-helper/SKILL.zh.md +14 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.en.md +18 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.md +39 -3
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks/SKILL.zh.md +16 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.en.md +14 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.md +54 -38
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-standalone-tasks-doc-review/SKILL.zh.md +14 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-story-assistant/SKILL.en.md +18 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-story-assistant/SKILL.md +37 -8
- package/node_modules/bmad-speckit/_bmad/cursor/skills/bmad-story-assistant/SKILL.zh.md +16 -1
- package/node_modules/bmad-speckit/_bmad/cursor/skills/speckit-workflow/SKILL.en.md +29 -0
- package/node_modules/bmad-speckit/_bmad/cursor/skills/speckit-workflow/SKILL.md +30 -2
- package/node_modules/bmad-speckit/_bmad/cursor/skills/speckit-workflow/SKILL.zh.md +30 -2
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/governance-packet-hard-closeout.cjs +172 -49
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/party-mode-read-current-session.cjs +1 -2
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/pre-continue-check.cjs +225 -1
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/runtime-policy-inject-core.cjs +536 -7
- package/node_modules/bmad-speckit/_bmad/skills/consumer-governance-validation/SKILL.md +30 -69
- package/node_modules/bmad-speckit/_bmad/skills/consumer-governance-validation/references/consumer-governance-validation.md +34 -258
- package/node_modules/bmad-speckit/package.json +6 -6
- package/node_modules/bmad-speckit/src/services/install-surface-manifest.js +52 -10
- package/node_modules/bmad-speckit/src/services/sync-service.js +5 -21
- package/node_modules/bmad-speckit/tests/pack-bmad-mirror.test.js +0 -2
- package/package.json +16 -12
- package/packages/bmad-speckit/node_modules/@bmad-speckit/ralph-method/package.json +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-context/package.json +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/auditor-post-actions.cjs +6 -0
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/emit-runtime-policy.cjs +2625 -408
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/resolve-for-session.cjs +3005 -909
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/run-auditor-host.cjs +259 -40
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/update-runtime-audit-index.cjs +6 -0
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/package.json +2 -7
- package/packages/bmad-speckit/node_modules/@bmad-speckit/schema/package.json +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/dist/dashboard/mcp-server.js +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.d.ts +2 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.d.ts.map +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.js +17 -2
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/dist/gate/version-lock.js.map +1 -1
- package/packages/bmad-speckit/node_modules/@bmad-speckit/scoring/package.json +1 -1
- package/scripts/assert-implementation-entry.ts +23 -37
- package/scripts/bmad-config.ts +13 -4
- package/scripts/bmad-help-routing-state.ts +149 -0
- package/scripts/bmad-runtime-worker.ts +89 -21
- package/scripts/continue-state-contract.ts +52 -0
- package/scripts/emit-runtime-policy.ts +142 -2
- package/scripts/governance-execution-result-ingestor.ts +13 -3
- package/scripts/governance-host-dispatch-adapter.ts +21 -0
- package/scripts/governance-packet-dispatch-worker.ts +7 -1
- package/scripts/governance-remediation-config.ts +12 -0
- package/scripts/governance-remediation-runner.ts +133 -13
- package/scripts/governance-runtime-queue.ts +2 -0
- package/scripts/i18n/resolve-for-session-cli.ts +5 -3
- package/scripts/init-to-root.js +9 -21
- package/scripts/live-smoke-main-agent-runtime.ts +309 -0
- package/scripts/live-smoke-speckit-workflow.ts +226 -0
- package/scripts/main-agent-orchestration.ts +1103 -0
- package/scripts/orchestration-dispatch-contract.ts +143 -0
- package/scripts/orchestration-state.ts +312 -0
- package/scripts/parse-bmad-audit-result.ts +7 -0
- package/scripts/prepublish-check.js +6 -7
- package/scripts/reviewer-schema.ts +26 -4
- package/scripts/run-auditor-host.ts +108 -5
- package/scripts/runtime-context-registry.ts +10 -0
- package/scripts/skill-inventory-provider.ts +0 -6
- package/scripts/skill-orchestration-audit.ts +336 -0
- package/scripts/speckit-cli.ts +1 -1
- package/scripts/validate-consumer-governance.ps1 +103 -134
- package/_bmad/runtime/hooks/governance-remediation-runner.cjs +0 -42
- package/_bmad/runtime/hooks/governance-runtime-worker.cjs +0 -43
- package/_bmad/runtime/hooks/run-bmad-runtime-worker.cjs +0 -1456
- package/node_modules/@bmad-speckit/runtime-emit/dist/governance-execution-result-ingestor.cjs +0 -10713
- package/node_modules/@bmad-speckit/runtime-emit/dist/governance-packet-dispatch-worker.cjs +0 -11055
- package/node_modules/@bmad-speckit/runtime-emit/dist/governance-packet-reconciler.cjs +0 -10712
- package/node_modules/@bmad-speckit/runtime-emit/dist/governance-remediation-runner.cjs +0 -16184
- package/node_modules/@bmad-speckit/runtime-emit/dist/governance-runtime-worker.cjs +0 -18563
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/governance-remediation-runner.cjs +0 -42
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/governance-runtime-worker.cjs +0 -43
- package/node_modules/bmad-speckit/_bmad/runtime/hooks/run-bmad-runtime-worker.cjs +0 -1456
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/governance-execution-result-ingestor.cjs +0 -10713
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/governance-packet-dispatch-worker.cjs +0 -11055
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/governance-packet-reconciler.cjs +0 -10712
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/governance-remediation-runner.cjs +0 -16184
- package/packages/bmad-speckit/node_modules/@bmad-speckit/runtime-emit/dist/governance-runtime-worker.cjs +0 -18563
package/README.md
CHANGED
|
@@ -46,10 +46,11 @@ Traditional AI tooling often stops at prompt orchestration. BMAD-Speckit-SDD-Flo
|
|
|
46
46
|
## Runtime Governance At A Glance
|
|
47
47
|
|
|
48
48
|
- **Four-Signal readiness gate** runs before implementation entry and keeps readiness scoring separate from implementation scoring.
|
|
49
|
-
-
|
|
50
|
-
-
|
|
51
|
-
- **
|
|
52
|
-
-
|
|
49
|
+
- **`main-agent-orchestration inspect` is the first read**: the main agent reads the repo-native authoritative surface before choosing the next global branch.
|
|
50
|
+
- **`dispatch-plan` is conditional**: only materialize a plan when the surface says the next branch should dispatch a packet.
|
|
51
|
+
- **Child agents only execute a bounded packet**: they return packet results, but they do not choose the next global branch.
|
|
52
|
+
- **`runAuditorHost` is post-audit close-out only**: it closes approved audits into scoring, dashboard, coach, and SFT outputs, then the main agent re-reads `inspect`.
|
|
53
|
+
- **Legacy worker/manual close-out playbooks are historical evidence**: keep them for auditability, but they are not the active accepted runtime path.
|
|
53
54
|
|
|
54
55
|
## Dashboard And MCP
|
|
55
56
|
|
|
@@ -164,16 +165,16 @@ This only removes installer-managed entries. It does not delete `.cursor`, `.cla
|
|
|
164
165
|
|
|
165
166
|
### Core Components
|
|
166
167
|
|
|
167
|
-
| Component | Purpose
|
|
168
|
-
| :-------------------------- |
|
|
169
|
-
| **`_bmad/`** | Canonical source of truth for workflow modules, hooks, prompts, routing, and host-facing assets.
|
|
170
|
-
| **`packages/scoring/`** | Scoring engine, readiness drift evaluation, dashboard projection, coach inputs, and SFT extraction.
|
|
171
|
-
| **`dashboard`** | Default runtime observability surface: live dashboard, runtime snapshots, and scoring projections.
|
|
172
|
-
| **`runtime-mcp`** | Optional MCP surface for agent tools over runtime data; enabled explicitly with `--with-mcp`.
|
|
173
|
-
| **`speckit-workflow`** | Specify -> Plan -> GAPS -> Tasks -> TDD with mandatory audit loops.
|
|
174
|
-
| **`bmad-story-assistant`** | Story lifecycle
|
|
175
|
-
| **`bmad-bug-assistant`** | Bug lifecycle path: RCA -> Party Mode -> BUGFIX -> Implement.
|
|
176
|
-
| **`bmad-standalone-tasks`** |
|
|
168
|
+
| Component | Purpose |
|
|
169
|
+
| :-------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------- |
|
|
170
|
+
| **`_bmad/`** | Canonical source of truth for workflow modules, hooks, prompts, routing, and host-facing assets. |
|
|
171
|
+
| **`packages/scoring/`** | Scoring engine, readiness drift evaluation, dashboard projection, coach inputs, and SFT extraction. |
|
|
172
|
+
| **`dashboard`** | Default runtime observability surface: live dashboard, runtime snapshots, and scoring projections. |
|
|
173
|
+
| **`runtime-mcp`** | Optional MCP surface for agent tools over runtime data; enabled explicitly with `--with-mcp`. |
|
|
174
|
+
| **`speckit-workflow`** | Specify -> Plan -> GAPS -> Tasks -> TDD with mandatory audit loops. |
|
|
175
|
+
| **`bmad-story-assistant`** | Story lifecycle entry: main agent reads `inspect`, dispatches bounded packets when needed, and closes post-audit through `runAuditorHost`. |
|
|
176
|
+
| **`bmad-bug-assistant`** | Bug lifecycle path: RCA -> Party Mode -> BUGFIX -> Implement, while the main Agent still owns the global `inspect -> dispatch-plan -> closeout` chain. |
|
|
177
|
+
| **`bmad-standalone-tasks`** | TASKS/BUGFIX execution still follows the main-agent path: `inspect` first, `dispatch-plan` only when needed, then bounded subagent work. |
|
|
177
178
|
|
|
178
179
|
<details>
|
|
179
180
|
<summary><b>View Folder Structure</b></summary>
|
|
@@ -197,6 +198,7 @@ BMAD-Speckit-SDD-Flow/
|
|
|
197
198
|
Key entry points:
|
|
198
199
|
|
|
199
200
|
- [Getting Started](docs/tutorials/getting-started.md)
|
|
201
|
+
- [Main-Agent Orchestration Reference](docs/reference/main-agent-orchestration.md)
|
|
200
202
|
- [Consumer Installation Guide](docs/how-to/consumer-installation.md)
|
|
201
203
|
- [Runtime Dashboard Guide](docs/how-to/runtime-dashboard.md)
|
|
202
204
|
- [Runtime MCP Installation](docs/how-to/runtime-mcp-installation.md)
|
package/README.zh-CN.md
CHANGED
|
@@ -46,10 +46,11 @@
|
|
|
46
46
|
## 运行时管控一览
|
|
47
47
|
|
|
48
48
|
- **四信号就绪检查**:在进入实现阶段前执行,与实现评分保持独立
|
|
49
|
-
-
|
|
50
|
-
-
|
|
51
|
-
-
|
|
52
|
-
-
|
|
49
|
+
- **先读 `main-agent-orchestration inspect`**:主 Agent 必须先读取 repo-native authoritative surface,再决定下一条全局分支
|
|
50
|
+
- **按需执行 `dispatch-plan`**:只有 surface 明确需要 materialize packet 时,才生成正式派发计划
|
|
51
|
+
- **子代理只执行 `bounded packet`**:子代理只返回 packet 结果,不负责决定下一条全局执行链
|
|
52
|
+
- **`runAuditorHost` 只负责 post-audit close-out**:审计通过后统一写入评分、看板、诊断和训练数据,然后主 Agent 重新读取 `inspect`
|
|
53
|
+
- **旧 worker / 手工 close-out 口径仅保留为历史证据**:可继续审计追溯,但不再是当前 accepted runtime path
|
|
53
54
|
|
|
54
55
|
## 看板与 MCP
|
|
55
56
|
|
|
@@ -164,16 +165,16 @@ npx --yes --package bmad-speckit-sdd-flow@latest bmad-speckit uninstall
|
|
|
164
165
|
|
|
165
166
|
### 核心组件
|
|
166
167
|
|
|
167
|
-
| 组件 | 说明
|
|
168
|
-
| :-------------------------- |
|
|
169
|
-
| **`_bmad/`** | 工作流模块、钩子、提示词、路由与宿主侧资产的规范源
|
|
170
|
-
| **`packages/scoring/`** | 评分引擎、就绪漂移评估、看板投影、诊断输入与训练数据提取
|
|
171
|
-
| **`dashboard`** | 默认运行时可观测层:实时看板、运行时快照、评分投影
|
|
172
|
-
| **`runtime-mcp`** | 可选的 MCP 工具接口,通过 `--with-mcp` 显式启用
|
|
173
|
-
| **`speckit-workflow`** | Specify → Plan → GAPS → Tasks → TDD,并带强制审计循环
|
|
174
|
-
| **`bmad-story-assistant`** | Story
|
|
175
|
-
| **`bmad-bug-assistant`** | Bug 生命周期路径:RCA → Party Mode → BUGFIX → Implement
|
|
176
|
-
| **`bmad-standalone-tasks`** | 针对 TASKS 或 BUGFIX
|
|
168
|
+
| 组件 | 说明 |
|
|
169
|
+
| :-------------------------- | :----------------------------------------------------------------------------------------------------------------- |
|
|
170
|
+
| **`_bmad/`** | 工作流模块、钩子、提示词、路由与宿主侧资产的规范源 |
|
|
171
|
+
| **`packages/scoring/`** | 评分引擎、就绪漂移评估、看板投影、诊断输入与训练数据提取 |
|
|
172
|
+
| **`dashboard`** | 默认运行时可观测层:实时看板、运行时快照、评分投影 |
|
|
173
|
+
| **`runtime-mcp`** | 可选的 MCP 工具接口,通过 `--with-mcp` 显式启用 |
|
|
174
|
+
| **`speckit-workflow`** | Specify → Plan → GAPS → Tasks → TDD,并带强制审计循环 |
|
|
175
|
+
| **`bmad-story-assistant`** | Story 生命周期入口:主 Agent 先读 `inspect`,按需派发 bounded packet,并在 post-audit 后通过 `runAuditorHost` 收口 |
|
|
176
|
+
| **`bmad-bug-assistant`** | Bug 生命周期路径:RCA → Party Mode → BUGFIX → Implement,但全局 `inspect -> dispatch-plan -> closeout` 主链仍由主 Agent 控制 |
|
|
177
|
+
| **`bmad-standalone-tasks`** | 针对 TASKS 或 BUGFIX 文档的执行仍先经过主 Agent `inspect`,必要时 `dispatch-plan`,再进入 bounded 子代理实施 |
|
|
177
178
|
|
|
178
179
|
<details>
|
|
179
180
|
<summary><b>查看目录结构</b></summary>
|
|
@@ -195,6 +196,7 @@ BMAD-Speckit-SDD-Flow/
|
|
|
195
196
|
## 文档入口
|
|
196
197
|
|
|
197
198
|
- [快速开始](docs/tutorials/getting-started.md)
|
|
199
|
+
- [主 Agent 编排参考](docs/reference/main-agent-orchestration.md)
|
|
198
200
|
- [消费项目安装指南](docs/how-to/consumer-installation.md)
|
|
199
201
|
- [运行时看板指南](docs/how-to/runtime-dashboard.md)
|
|
200
202
|
- [运行时 MCP 安装](docs/how-to/runtime-mcp-installation.md)
|
|
@@ -4,6 +4,32 @@ packetHosts:
|
|
|
4
4
|
- cursor
|
|
5
5
|
- claude
|
|
6
6
|
- codex
|
|
7
|
+
execution:
|
|
8
|
+
enabled: true
|
|
9
|
+
interactiveMode: main-agent
|
|
10
|
+
fallbackAutonomousMode: false
|
|
11
|
+
authoritativeHost: cursor
|
|
12
|
+
fallbackHosts:
|
|
13
|
+
- claude
|
|
14
|
+
- codex
|
|
15
|
+
dispatch:
|
|
16
|
+
leaseTimeoutSeconds: 900
|
|
17
|
+
heartbeatIntervalSeconds: 60
|
|
18
|
+
maxDispatchAttempts: 3
|
|
19
|
+
execution:
|
|
20
|
+
timeoutMinutes: 30
|
|
21
|
+
maxExecutionAttempts: 2
|
|
22
|
+
rerunGate:
|
|
23
|
+
required: true
|
|
24
|
+
autoSchedule: true
|
|
25
|
+
maxGateRetries: 2
|
|
26
|
+
escalation:
|
|
27
|
+
afterDispatchFailures: 3
|
|
28
|
+
afterExecutionFailures: 2
|
|
29
|
+
afterGateFailures: 2
|
|
30
|
+
projections:
|
|
31
|
+
emitNonAuthoritativePackets: true
|
|
32
|
+
archivePath: _bmad-output/runtime/governance/archive
|
|
7
33
|
provider:
|
|
8
34
|
mode: openai-compatible
|
|
9
35
|
id: openai-governance
|
|
@@ -70,6 +70,8 @@ BUGFIX 文档严格审计 Agent,面向 bugfix 流程中的 BUGFIX 文档与修
|
|
|
70
70
|
- `convergence_status`
|
|
71
71
|
- `next_action`
|
|
72
72
|
- `ready`
|
|
73
|
+
- `mainAgentNextAction`
|
|
74
|
+
- `mainAgentReady`
|
|
73
75
|
|
|
74
76
|
## Mandatory Startup
|
|
75
77
|
|
|
@@ -141,8 +143,12 @@ handoff:
|
|
|
141
143
|
next_action: revise_bugfix_doc|implement_bugfix
|
|
142
144
|
next_agent: auditor-bugfix|bmad-bug-agent
|
|
143
145
|
ready: true|false
|
|
146
|
+
mainAgentNextAction: dispatch_remediation|dispatch_implement
|
|
147
|
+
mainAgentReady: true|false
|
|
144
148
|
```
|
|
145
149
|
|
|
150
|
+
说明:这里的 `next_agent`、`mainAgentNextAction`、`mainAgentReady` 只作为 compatibility hint。审计返回后,真正的下一条全局分支必须回到主 Agent,并重新读取 `main-agent-orchestration` surface 后决定。
|
|
151
|
+
|
|
146
152
|
## Lifecycle
|
|
147
153
|
|
|
148
154
|
1. Pre-Audit
|
|
@@ -141,7 +141,7 @@ Gap 修复决策:
|
|
|
141
141
|
|
|
142
142
|
- `artifactDocPath`: 被审 spec.md 文件路径(必填)
|
|
143
143
|
- `reportPath`: 审计报告保存路径(必填)
|
|
144
|
-
- `storyPath`: 原始 Story
|
|
144
|
+
- `storyPath`: 原始 Story 文档路径(**Story-flow 下必填**;standalone speckit 可不传)
|
|
145
145
|
- `prdPath`: PRD 文档路径(可选)
|
|
146
146
|
- `archPath`: 架构文档路径(可选)
|
|
147
147
|
- `epic`: Epic 编号
|
|
@@ -157,9 +157,15 @@ Gap 修复决策:
|
|
|
157
157
|
2. **读取批判审计员规范**:`.claude/skills/speckit-workflow/references/audit-prompts-critical-auditor-appendix.md`
|
|
158
158
|
3. **读取文档迭代规则**:`.claude/skills/speckit-workflow/references/audit-document-iteration-rules.md`
|
|
159
159
|
4. **读取被审文档**:`artifactDocPath` 指定的spec.md
|
|
160
|
-
5. **读取原始需求文档**:Story/PRD/ARCH
|
|
160
|
+
5. **读取原始需求文档**:Story/PRD/ARCH(如果提供路径;若当前为 Story-flow,则必须读取 `storyPath`)
|
|
161
161
|
6. **检查前置审计标记**:查看spec.md 末尾是否有`<!-- AUDIT: PASSED` 标记
|
|
162
162
|
|
|
163
|
+
### Story-flow 硬约束
|
|
164
|
+
|
|
165
|
+
- 若 `artifactDocPath` 属于 Story-flow `spec-E{epic}-S{story}.md`,则 `storyPath` 为**强制输入**,不得省略。
|
|
166
|
+
- 审计报告控制字段中必须显式回传 `storyPath: <path>`,供 `runAuditorHost` 执行 `Story -> Spec` 版本锁。
|
|
167
|
+
- 缺失 `storyPath` 时,结论不得视为可 closeout。
|
|
168
|
+
|
|
163
169
|
## Execution Flow
|
|
164
170
|
|
|
165
171
|
### Step 1: 模型选择信息输出
|
|
@@ -70,6 +70,8 @@
|
|
|
70
70
|
- `convergence_status`
|
|
71
71
|
- `next_action`
|
|
72
72
|
- `ready`
|
|
73
|
+
- `mainAgentNextAction`
|
|
74
|
+
- `mainAgentReady`
|
|
73
75
|
|
|
74
76
|
## Mandatory Startup
|
|
75
77
|
|
|
@@ -141,8 +143,12 @@ handoff:
|
|
|
141
143
|
next_action: revise_tasks_doc|execute_standalone_tasks
|
|
142
144
|
next_agent: auditor-tasks-doc|bmad-standalone-tasks
|
|
143
145
|
ready: true|false
|
|
146
|
+
mainAgentNextAction: dispatch_remediation|dispatch_implement
|
|
147
|
+
mainAgentReady: true|false
|
|
144
148
|
```
|
|
145
149
|
|
|
150
|
+
说明:这里的 `next_agent`、`mainAgentNextAction`、`mainAgentReady` 只作为 compatibility hint。审计返回后,真正的下一条全局分支必须回到主 Agent,并重新读取 `main-agent-orchestration` surface 后决定。
|
|
151
|
+
|
|
146
152
|
## Lifecycle
|
|
147
153
|
|
|
148
154
|
1. Pre-Audit
|
|
@@ -274,9 +274,10 @@ PROMPT_PATH="_bmad-output/epic-{epic}-{epic-slug}/PROMPT_audit-epic-{epic}_round
|
|
|
274
274
|
#### Step 3.2: 调用审计 Agent
|
|
275
275
|
|
|
276
276
|
**Primary Executor**: `auditor-epic`
|
|
277
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
277
278
|
|
|
278
279
|
```typescript
|
|
279
|
-
|
|
280
|
+
MainAgentDispatchRequest({
|
|
280
281
|
description: "审计 Epic {epic} 完成状态",
|
|
281
282
|
subagent_type: "general-purpose",
|
|
282
283
|
prompt: `
|
|
@@ -392,6 +393,8 @@ completion_report: "_bmad-output/epic-E001-test-epic/EPIC_COMPLETION_REPORT.md"
|
|
|
392
393
|
next_action: "proceed_to_layer_5" # 进入收尾层
|
|
393
394
|
```
|
|
394
395
|
|
|
396
|
+
说明:该 handoff 只作为 compatibility hint。Epic 审计后是否进入下一层,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
397
|
+
|
|
395
398
|
## Constraints
|
|
396
399
|
|
|
397
400
|
- **禁止自行 commit**
|
|
@@ -34,15 +34,27 @@ BMAD Speckit SDD 流程的总控 Agent,负责多 Story 管理、阶段门控
|
|
|
34
34
|
- `.claude/protocols/audit-result-schema.md`
|
|
35
35
|
- `.claude/protocols/handoff-schema.md`
|
|
36
36
|
- `.claude/protocols/commit-protocol.md`
|
|
37
|
+
7. Read repo-native orchestration surface:
|
|
38
|
+
- `npm run main-agent-orchestration -- --cwd {project-root} --action inspect`
|
|
39
|
+
- if dispatch is needed: `--action dispatch-plan|claim|dispatch|complete|invalidate`
|
|
37
40
|
|
|
38
41
|
### Auto-Continue Detection
|
|
39
42
|
|
|
40
43
|
`bmad-master` 必须区分:
|
|
41
44
|
|
|
42
|
-
1. **handoff
|
|
43
|
-
2. **
|
|
45
|
+
1. **stage handoff semantics**:story state / handoff 中可保留 `next_action` 与 `ready`,用于表达子阶段推荐结果
|
|
46
|
+
2. **authoritative main-agent readiness**:优先读取 repo-native `main-agent-orchestration` surface;若该 surface 不可用,再读取 story state / handoff 中的 `mainAgentNextAction` 与 `mainAgentReady: true`
|
|
47
|
+
3. **runtime permission to continue**:当前运行时是否显式启用了 auto-continue(CLI `--continue`、环境变量 `BMAD_AUTO_CONTINUE=true`、或配置文件 `auto_continue.enabled: true`)
|
|
44
48
|
|
|
45
|
-
|
|
49
|
+
只有 2 与 3 同时满足时,才允许自动推进;若 repo-native `main-agent-orchestration` surface 与 `mainAgentNextAction / mainAgentReady` 冲突,以前者为准;若 `next_action` / `ready` 与 `mainAgentNextAction / mainAgentReady` 冲突,以后者为准。**但当 repo-native surface 可用时,不得直接从 handoff 或 story state 派发下一执行体。**
|
|
50
|
+
|
|
51
|
+
### Canonical Runtime Surface
|
|
52
|
+
|
|
53
|
+
主 Agent 的正式消费顺序是:
|
|
54
|
+
|
|
55
|
+
1. `main-agent-orchestration`
|
|
56
|
+
2. `mainAgentNextAction / mainAgentReady`
|
|
57
|
+
3. `next_action / ready`
|
|
46
58
|
|
|
47
59
|
```yaml
|
|
48
60
|
auto_continue_check:
|
|
@@ -54,36 +66,39 @@ auto_continue_check:
|
|
|
54
66
|
|
|
55
67
|
conditions:
|
|
56
68
|
- auto_continue.enabled === true
|
|
57
|
-
-
|
|
58
|
-
-
|
|
69
|
+
- orchestration_surface.mainAgentNextAction 存在
|
|
70
|
+
- orchestration_surface.mainAgentReady === true
|
|
71
|
+
- orchestration_surface.source in [orchestration_state, implementation_entry_gate, reviewer_closeout]
|
|
59
72
|
|
|
60
73
|
action:
|
|
61
|
-
- 自动解析
|
|
74
|
+
- 自动解析 mainAgentNextAction 对应的主控分支
|
|
75
|
+
- 如需派发,先执行 dispatch-plan / claim / dispatch
|
|
62
76
|
- 更新 story state stage / follow_up
|
|
63
77
|
- 路由到下一个执行体
|
|
64
78
|
- 输出:
|
|
65
79
|
auto_proceed: true
|
|
66
|
-
reason: "auto_continue enabled 且
|
|
80
|
+
reason: "auto_continue enabled 且 orchestration_surface.mainAgentNextAction={action} 且 mainAgentReady=true"
|
|
67
81
|
```
|
|
68
82
|
|
|
69
|
-
若
|
|
83
|
+
若 repo-native orchestration surface 已就绪,但 `auto_continue.enabled !== true`,则:
|
|
70
84
|
- 不自动推进
|
|
71
85
|
- 仅输出建议下一步
|
|
72
86
|
- 等待用户确认或显式用 `--continue` 重入
|
|
73
87
|
|
|
74
|
-
|
|
88
|
+
若 repo-native orchestration surface 不可用,且仅 handoff 中的 `mainAgentNextAction / mainAgentReady` 已就绪,则:
|
|
89
|
+
- 这只属于 compatibility fallback
|
|
90
|
+
- 不得据此绕过 surface-first 规则重定义 accepted runtime path
|
|
91
|
+
- 仅在兼容恢复 / 旧 story state 回放时使用
|
|
92
|
+
|
|
93
|
+
**mainAgentNextAction 到主控分支的映射**:
|
|
75
94
|
|
|
76
95
|
```yaml
|
|
77
|
-
|
|
78
|
-
|
|
79
|
-
|
|
80
|
-
|
|
81
|
-
|
|
82
|
-
|
|
83
|
-
implement: implement_passed
|
|
84
|
-
document_audit: document_audit_passed
|
|
85
|
-
commit_gate: commit_gate_passed
|
|
86
|
-
commit: commit_ready
|
|
96
|
+
main_agent_action_map:
|
|
97
|
+
dispatch_review: route_to_auditor_or_reviewer
|
|
98
|
+
dispatch_implement: route_to_implementation_executor
|
|
99
|
+
dispatch_remediation: rerun_fix_or_document_loop
|
|
100
|
+
run_closeout: enter_closeout_or_commit_gate
|
|
101
|
+
await_user: stop_and_wait_for_user
|
|
87
102
|
```
|
|
88
103
|
|
|
89
104
|
**Output 增强**:
|
|
@@ -91,15 +106,15 @@ action_to_stage_map:
|
|
|
91
106
|
当 auto-continue 触发时,输出:
|
|
92
107
|
```yaml
|
|
93
108
|
auto_proceed: true
|
|
94
|
-
reason: "auto_continue enabled 且
|
|
95
|
-
follow_up: <next_agent>
|
|
109
|
+
reason: "auto_continue enabled 且 orchestration_surface.mainAgentNextAction={action} 且 mainAgentReady=true"
|
|
110
|
+
follow_up: <next_agent|dispatch>
|
|
96
111
|
```
|
|
97
112
|
|
|
98
113
|
当 handoff 已就绪但未开启 auto-continue 时,输出:
|
|
99
114
|
```yaml
|
|
100
115
|
auto_proceed: false
|
|
101
|
-
reason: "handoff 已就绪,但未启用 --continue / BMAD_AUTO_CONTINUE / auto_continue.enabled"
|
|
102
|
-
follow_up: <next_agent>
|
|
116
|
+
reason: "main-agent handoff 已就绪,但未启用 --continue / BMAD_AUTO_CONTINUE / auto_continue.enabled"
|
|
117
|
+
follow_up: <next_agent|dispatch>
|
|
103
118
|
suggested_command: "@bmad-master 继续 {epic}-{story} --continue"
|
|
104
119
|
```
|
|
105
120
|
|
|
@@ -500,7 +515,7 @@ Always state:
|
|
|
500
515
|
4. `denial_reason` (if deny)
|
|
501
516
|
5. `follow_up`: next agent/action
|
|
502
517
|
6. `state_patch`: 应写回 story state 的变更
|
|
503
|
-
7. `auto_proceed`: true (当
|
|
518
|
+
7. `auto_proceed`: true (当 mainAgentNextAction 存在且 mainAgentReady=true 时)
|
|
504
519
|
8. `reason`: auto_proceed 触发原因
|
|
505
520
|
|
|
506
521
|
## Example Usage
|
|
@@ -201,7 +201,7 @@ Claude 端 Stage 2 Story 审计执行体,负责审计 Story 文档并决定是
|
|
|
201
201
|
|
|
202
202
|
### Runtime Contracts
|
|
203
203
|
- 审计报告路径:`_bmad-output/implementation-artifacts/epic-{epic_num}-{epic_slug}/story-{story_num}-{story_slug}/AUDIT_story-{epic_num}-{story_num}.md`(与 create-story workflow、bmad-story-assistant 一致:story 子目录为 `story-{story}-{slug}`)
|
|
204
|
-
- 审计通过:更新 story state 为 `story_audit_passed
|
|
204
|
+
- 审计通过:更新 story state 为 `story_audit_passed`;`handoff -> speckit-implement` 只作为 compatibility hint,真正是否进入 Dev Story 必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration` surface 后决定
|
|
205
205
|
- 审计失败:更新 story state 为 `story_audit_failed`,要求修 Story 文档后重新审计
|
|
206
206
|
|
|
207
207
|
## Repo Add-ons
|
|
@@ -223,6 +223,8 @@ next_action: dev_story
|
|
|
223
223
|
next_agent: speckit-implement
|
|
224
224
|
```
|
|
225
225
|
|
|
226
|
+
说明:`next_agent: speckit-implement` 只作为 compatibility hint。
|
|
227
|
+
|
|
226
228
|
### FAIL
|
|
227
229
|
```yaml
|
|
228
230
|
layer: 3
|
|
@@ -232,3 +234,5 @@ auditReportPath: _bmad-output/implementation-artifacts/epic-{epic_num}-{epic_slu
|
|
|
232
234
|
next_action: revise_story
|
|
233
235
|
next_agent: bmad-story-create
|
|
234
236
|
```
|
|
237
|
+
|
|
238
|
+
说明:`next_agent: bmad-story-create` 同样只作为 compatibility hint。
|
|
@@ -159,7 +159,7 @@ Claude 端 Stage 1 Create Story 执行体,负责在 BMAD Story 流程中生成
|
|
|
159
159
|
### Runtime Contracts
|
|
160
160
|
- 产物路径:`_bmad-output/implementation-artifacts/epic-{epic_num}-{epic_slug}/story-{story_num}-{story_slug}/{epic_num}-{story_num}-{story_slug}.md`(与 create-story workflow、bmad-story-assistant 一致:story 子目录为 `story-{story}-{slug}`,不含 epic 编号)
|
|
161
161
|
- Story 产出完成后,必须将 story state 更新为 `story_created`
|
|
162
|
-
- 必须写入 handoff
|
|
162
|
+
- 必须写入 handoff;`next_agent: bmad-story-audit` 只作为 compatibility hint,真正是否进入 Stage 2 必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration` surface 后决定
|
|
163
163
|
- 若用户明确跳过 Create Story,必须记录跳过依据并直接进入 Story 审计
|
|
164
164
|
|
|
165
165
|
## Repo Add-ons
|
|
@@ -180,3 +180,5 @@ artifactPath: _bmad-output/implementation-artifacts/epic-{epic_num}-{epic_slug}/
|
|
|
180
180
|
next_action: story_audit
|
|
181
181
|
next_agent: bmad-story-audit
|
|
182
182
|
```
|
|
183
|
+
|
|
184
|
+
说明:本执行体不得直接把控制链交给 `bmad-story-audit`;handoff 只是给主 Agent 的兼容提示。
|
|
@@ -56,8 +56,9 @@ Bug 修复专用流程。
|
|
|
56
56
|
- BUGFIX 文档审计未执行、未通过或结论不明时,一律视为实现前阻断
|
|
57
57
|
- 修复完成后必须进行实施后审计
|
|
58
58
|
- 提交仅允许通过 `bmad-master` 的 `commit_request`
|
|
59
|
-
- 返回必须包含:`execution_summary`、`artifacts`、`handoff`、`next_action`、`ready`
|
|
59
|
+
- 返回必须包含:`execution_summary`、`artifacts`、`handoff`、`next_action`、`ready`、`mainAgentNextAction`、`mainAgentReady`
|
|
60
60
|
- 修复流程必须保留 RCA / BUGFIX / repro test / post-audit 的完整链路证据
|
|
61
|
+
- parent 主 Agent 在发起或回收本执行体前,必须优先读取 `npm run main-agent-orchestration -- --cwd {project-root} --action inspect|dispatch-plan`
|
|
61
62
|
|
|
62
63
|
## Repo Add-ons
|
|
63
64
|
|
|
@@ -109,6 +110,7 @@ Bug 修复专用流程。
|
|
|
109
110
|
|
|
110
111
|
- 主 Agent 调用执行体时,必须完整复制 Cursor 侧 bug assistant / bugfix workflow 的对应正文模板,不得摘要化
|
|
111
112
|
- 不得只传文件路径让执行体自行推断规则
|
|
113
|
+
- parent 主 Agent 不得只根据 `mainAgentNextAction / mainAgentReady` 派发本执行体;若 repo-native orchestration surface 可用,必须优先消费该 surface
|
|
112
114
|
- 必须显式传入:
|
|
113
115
|
- bug 描述
|
|
114
116
|
- RCA 摘要或 RCA 文档路径
|
|
@@ -130,8 +132,12 @@ handoff:
|
|
|
130
132
|
next_action: implement_bugfix|post_audit|commit_gate|revise_bugfix_doc
|
|
131
133
|
next_agent: bmad-bug-agent|auditor-implement|bmad-master|auditor-bugfix
|
|
132
134
|
ready: true|false
|
|
135
|
+
mainAgentNextAction: dispatch_implement|dispatch_review|dispatch_remediation|run_closeout|await_user
|
|
136
|
+
mainAgentReady: true|false
|
|
133
137
|
```
|
|
134
138
|
|
|
139
|
+
说明:这里的 `next_agent`、`mainAgentNextAction`、`mainAgentReady` 只作为 compatibility hint。bugfix 子执行体返回后,主 Agent 必须重新读取 `main-agent-orchestration` surface,再决定下一条全局分支。
|
|
140
|
+
|
|
135
141
|
## State Updates
|
|
136
142
|
|
|
137
143
|
```yaml
|
|
@@ -136,8 +136,12 @@ handoff:
|
|
|
136
136
|
next_action: scoring_trigger|return_to_auditor
|
|
137
137
|
next_agent: bmad-master|auditor-implement
|
|
138
138
|
ready: true|false
|
|
139
|
+
mainAgentNextAction: dispatch_review
|
|
140
|
+
mainAgentReady: true|false
|
|
139
141
|
```
|
|
140
142
|
|
|
143
|
+
说明:这里的 `next_agent`、`mainAgentNextAction`、`mainAgentReady` 只作为 compatibility hint。review 返回后,主 Agent 必须重新读取 `main-agent-orchestration` surface,再决定下一条全局分支。
|
|
144
|
+
|
|
141
145
|
## State Updates
|
|
142
146
|
|
|
143
147
|
```yaml
|
|
@@ -243,9 +243,10 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
243
243
|
#### Step 3.2: 调用审计 Agent
|
|
244
244
|
|
|
245
245
|
**Primary Executor**: `auditor-gaps` 通过 `subagent_type: general-purpose` 调用
|
|
246
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
246
247
|
|
|
247
248
|
```typescript
|
|
248
|
-
|
|
249
|
+
MainAgentDispatchRequest({
|
|
249
250
|
description: "审计 IMPLEMENTATION_GAPS-E{epic}-S{story}.md",
|
|
250
251
|
subagent_type: "general-purpose",
|
|
251
252
|
prompt: `
|
|
@@ -337,6 +338,8 @@ auditReportPath: specs/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/AUDIT_
|
|
|
337
338
|
next_action: proceed_to_tasks
|
|
338
339
|
```
|
|
339
340
|
|
|
341
|
+
说明:handoff 只作为 compatibility hint;是否进入 `tasks`,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
342
|
+
|
|
340
343
|
## Constraints
|
|
341
344
|
|
|
342
345
|
- **禁止自行 commit**
|
|
@@ -420,9 +420,10 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
420
420
|
#### Step 5.2: 调用审计 Agent
|
|
421
421
|
|
|
422
422
|
**Primary Executor**: `auditor-implement` 通过 `subagent_type: general-purpose` 调用
|
|
423
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
423
424
|
|
|
424
425
|
```typescript
|
|
425
|
-
|
|
426
|
+
MainAgentDispatchRequest({
|
|
426
427
|
description: "审计 implement 阶段代码实现",
|
|
427
428
|
subagent_type: "general-purpose",
|
|
428
429
|
prompt: `
|
|
@@ -547,6 +548,8 @@ tddSummary:
|
|
|
547
548
|
next_action: commit_gate
|
|
548
549
|
```
|
|
549
550
|
|
|
551
|
+
说明:handoff 只作为 compatibility hint;是否进入 closeout / commit gate,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
552
|
+
|
|
550
553
|
## Constraints
|
|
551
554
|
|
|
552
555
|
- **禁止自行 commit**
|
|
@@ -253,9 +253,10 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
253
253
|
#### Step 4.2: 调用审计 Agent
|
|
254
254
|
|
|
255
255
|
**Primary Executor**: `auditor-plan` 通过 `subagent_type: general-purpose` 调用
|
|
256
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
256
257
|
|
|
257
258
|
```typescript
|
|
258
|
-
|
|
259
|
+
MainAgentDispatchRequest({
|
|
259
260
|
description: "审计 plan-E{epic}-S{story}.md",
|
|
260
261
|
subagent_type: "general-purpose",
|
|
261
262
|
prompt: `
|
|
@@ -363,6 +364,8 @@ auditReportPath: specs/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/AUDIT_
|
|
|
363
364
|
next_action: proceed_to_gaps
|
|
364
365
|
```
|
|
365
366
|
|
|
367
|
+
说明:handoff 只作为 compatibility hint;是否进入 `gaps`,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
368
|
+
|
|
366
369
|
## Constraints
|
|
367
370
|
|
|
368
371
|
- **禁止自行 commit**
|
|
@@ -228,6 +228,9 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
228
228
|
- `_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/PROMPT_audit-spec-E{epic}-S{story}_round{N}.md`
|
|
229
229
|
- 审计报告输出路径:
|
|
230
230
|
- `specs/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/AUDIT_spec-E{epic}-S{story}.md`
|
|
231
|
+
- Story-flow 约束:
|
|
232
|
+
- 若当前产物为 `spec-E{epic}-S{story}.md`,则审计子任务必须显式携带 `storyPath`
|
|
233
|
+
- 审计报告控制字段必须回传 `storyPath: <Story 文档路径>`,供 `runAuditorHost` 执行 `Story -> Spec` source_hash 版本锁
|
|
231
234
|
- 审计失败处理:
|
|
232
235
|
- 主 Agent 根据 required_fixes 修改 spec 文档后重新发起下一轮审计- 审计通过处理:
|
|
233
236
|
- 追加通过标记
|
|
@@ -260,9 +263,10 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
260
263
|
|
|
261
264
|
**子任务prompt 不得再只是"读取并执行某个 Prompt 文件"。**
|
|
262
265
|
**必须保证子任务收到的主文本基线仍然是 Cursor §1,而不是无法区分来源的本地重写版。**
|
|
266
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
263
267
|
|
|
264
268
|
```typescript
|
|
265
|
-
|
|
269
|
+
MainAgentDispatchRequest({
|
|
266
270
|
description: "审计spec-E{epic}-S{story}.md",
|
|
267
271
|
subagent_type: "general-purpose",
|
|
268
272
|
prompt: `
|
|
@@ -275,6 +279,8 @@ Task({
|
|
|
275
279
|
**Claude/OMC Runtime Adapter**
|
|
276
280
|
- 审计报告输出至
|
|
277
281
|
specs/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/AUDIT_spec-E{epic}-S{story}.md
|
|
282
|
+
- Story-flow 必须携带
|
|
283
|
+
- `storyPath: _bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/{epic}-{story}-{story-slug}.md`
|
|
278
284
|
- 同时保存本轮 Prompt 存档至
|
|
279
285
|
_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/PROMPT_audit-spec-E{epic}-S{story}_round{N}.md
|
|
280
286
|
|
|
@@ -363,6 +369,8 @@ auditReportPath: specs/epic-{epic}-{epic-slug}/story-{story}-{story-slug}/AUDIT_
|
|
|
363
369
|
next_action: proceed_to_plan
|
|
364
370
|
```
|
|
365
371
|
|
|
372
|
+
说明:handoff 只作为 compatibility hint;是否进入 `plan`,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
373
|
+
|
|
366
374
|
## Constraints
|
|
367
375
|
|
|
368
376
|
- **禁止自行 commit**
|
|
@@ -320,9 +320,10 @@ PROMPT_PATH="_bmad-output/implementation-artifacts/epic-{epic}-{epic-slug}/story
|
|
|
320
320
|
#### Step 4.2: 调用审计 Agent
|
|
321
321
|
|
|
322
322
|
**Primary Executor**: `auditor-tasks` 通过 `subagent_type: general-purpose` 调用
|
|
323
|
+
**当前 accepted path(强制)**:本执行体不得在自身上下文里直接 `Task(...)` 派发 auditor;以下内容只能作为返回给主 Agent 的 dispatch request / compatibility hint。真实派发必须回到主 Agent,由主 Agent 重新读取 `main-agent-orchestration inspect`,必要时执行 `dispatch-plan`,再决定是否派发 auditor。
|
|
323
324
|
|
|
324
325
|
```typescript
|
|
325
|
-
|
|
326
|
+
MainAgentDispatchRequest({
|
|
326
327
|
description: "审计 tasks-E{epic}-S{story}.md",
|
|
327
328
|
subagent_type: "general-purpose",
|
|
328
329
|
prompt: `
|
|
@@ -421,6 +422,8 @@ next_action: proceed_to_implement
|
|
|
421
422
|
next_agent: speckit-implement
|
|
422
423
|
```
|
|
423
424
|
|
|
425
|
+
说明:`next_agent: speckit-implement` 只作为 compatibility hint;是否进入 implement,必须回到主 Agent,由主 Agent 重新读取 authoritative surface 后决定。
|
|
426
|
+
|
|
424
427
|
## Constraints
|
|
425
428
|
|
|
426
429
|
- **禁止自行 commit**
|
|
@@ -52,8 +52,9 @@
|
|
|
52
52
|
- 前置门控未执行、未通过或结论不明时,一律视为实施前阻断
|
|
53
53
|
- 批次门控:每批任务完成后调用实现审计
|
|
54
54
|
- 提交门控:仅允许向 `bmad-master` 返回 `commit_request`,禁止自行 commit
|
|
55
|
-
- 返回必须包含:`execution_summary`、`artifacts`、`handoff`、`next_action`、`ready`
|
|
55
|
+
- 返回必须包含:`execution_summary`、`artifacts`、`handoff`、`next_action`、`ready`、`mainAgentNextAction`、`mainAgentReady`
|
|
56
56
|
- 实施过程必须维护 standalone 模式下的 prd/progress 产物
|
|
57
|
+
- parent 主 Agent 在发起或回收本执行体前,必须优先读取 `npm run main-agent-orchestration -- --cwd {project-root} --action inspect|dispatch-plan`
|
|
57
58
|
|
|
58
59
|
## Repo Add-ons
|
|
59
60
|
|
|
@@ -112,6 +113,7 @@
|
|
|
112
113
|
|
|
113
114
|
- 主 Agent 调用实施子代理时,必须完整复制 Cursor 侧对应正文模板,不得摘要化
|
|
114
115
|
- 不得只传文件路径让子代理自行推断规则
|
|
116
|
+
- parent 主 Agent 不得只根据 `mainAgentNextAction / mainAgentReady` 派发本执行体;若 repo-native orchestration surface 可用,必须优先消费该 surface
|
|
115
117
|
- 必须显式传入:
|
|
116
118
|
- 文档路径
|
|
117
119
|
- 未完成任务清单
|
|
@@ -134,8 +136,12 @@ handoff:
|
|
|
134
136
|
next_action: implement_next_batch|post_batch_audit|commit_gate|revise_tasks_doc
|
|
135
137
|
next_agent: bmad-standalone-tasks|auditor-implement|bmad-master|auditor-tasks-doc
|
|
136
138
|
ready: true|false
|
|
139
|
+
mainAgentNextAction: dispatch_implement|dispatch_review|dispatch_remediation|run_closeout|await_user
|
|
140
|
+
mainAgentReady: true|false
|
|
137
141
|
```
|
|
138
142
|
|
|
143
|
+
说明:这里的 `next_agent`、`mainAgentNextAction`、`mainAgentReady` 只作为 compatibility hint。standalone 执行体返回后,主 Agent 必须重新读取 `main-agent-orchestration` surface,再决定下一条全局分支。
|
|
144
|
+
|
|
139
145
|
## State Updates
|
|
140
146
|
|
|
141
147
|
```yaml
|