speccrew 0.7.74 → 0.7.75

package/.speccrew/skills/speccrew-knowledge-bizs-ui-style-extract/SKILL.md

@@ -4,399 +4,12 @@ description: Extract and aggregate UI design patterns (page types, components, l
 tools: Read, Write
 ---
 
-# Bizs UI Style Extract (XML Workflow)
-
-Extract and aggregate **UI design patterns** from bizs pipeline analyzed feature documents using XML workflow blocks. Through cross-module clustering analysis, identify common page types, component patterns, and layout patterns, then output them to the techs knowledge base `ui-style-patterns/` subdirectory.
-
-## Language Adaptation
-
-**CRITICAL**: All generated documents must match the user's language. Detect the language from the user's input and generate content accordingly.
-
-- User writes in 中文 → Generate Chinese documents, use `language: "zh"`
-- User writes in English → Generate English documents, use `language: "en"`
-- User writes in other languages → Use appropriate language code
-
-**All generated pattern documents must be in the language specified by the `language` parameter.**
-
-## Trigger Scenarios
-
-- Called by `speccrew-knowledge-bizs-dispatch` Stage 3.5 (after Module Summarize, before System Summary)
+# Trigger Scenarios
+- Called by speccrew-knowledge-bizs-dispatch Stage 3.5
 - "Extract UI patterns from bizs features"
 - "Aggregate UI design patterns"
 
-## Input
-
-| Variable | Description | Required |
-|----------|-------------|----------|
-| `platform_id` | Platform identifier (e.g., web-vue, mobile-uniapp), used to locate output directory | Yes |
-| `platform_type` | Platform type (web, mobile, desktop), only execute for frontend platforms | Yes |
-| `feature_docs_path` | Completed feature documents base path, e.g., `speccrew-workspace/knowledges/bizs/{platform-type}/{module}/features/` | Yes |
-| `features_manifest_path` | Path to features-{platform}.json, used to get completed feature list | Yes |
-| `module_overviews_path` | **Parent directory** containing all module overview subdirectories. Example: `knowledges/bizs/web-vue/` (this directory contains `system/system-overview.md`, `user/user-overview.md`, etc.). **NOT** a specific module directory like `knowledges/bizs/web-vue/system/`. | Yes |
-| `output_path` | Output directory, e.g., `speccrew-workspace/knowledges/techs/{platform_id}/ui-style-patterns/` | Yes |
-| `language` | User language code | Yes |
-
-## Output
-
-> **Directory Separation**: This skill outputs to `ui-style-patterns/` (NOT `ui-style/`).
-> - `ui-style/` is managed by techs pipeline (framework-level design system, existing components/pages)
-> - `ui-style-patterns/` is managed by bizs pipeline (business pattern aggregation from feature docs)
-> This separation prevents file conflicts between the two pipelines.
-
-```
-{output_path}/
-├── page-types/          # Page type pattern documents
-│   ├── {pattern-name}.md
-│   └── ...
-├── components/          # Component pattern documents
-│   ├── {pattern-name}.md
-│   └── ...
-└── layouts/             # Layout pattern documents
-    ├── {pattern-name}.md
-    └── ...
-```
-
-## Absolute Constraints
-
-> **These rules apply to ALL document generation steps. Violation = task failure.**
-
-1. **FORBIDDEN: `create_file` for pattern documents** — NEVER use `create_file` to write pattern documents. Each document MUST be created by copying the appropriate template then filling sections with `search_replace`.
-
-2. **FORBIDDEN: Full-file rewrite** — NEVER replace the entire document content in a single operation. Always use targeted `search_replace` on specific sections.
-
-3. **MANDATORY: Template-first workflow** — Copy template MUST execute before filling sections for every pattern document.
-
 ## AgentFlow Definition
-
 <!-- @agentflow: SKILL.xml -->
-
-> **REQUIRED**: Before executing this workflow, read the XML workflow specification: `speccrew-workspace/docs/rules/agentflow-spec.md`
-
-  <!-- ============================================================
-       Input Parameters Definition
-       ============================================================ -->
-  <block type="input" id="I1" desc="Workflow input parameters">
-    <field name="platform_id" required="true" type="string" desc="Platform identifier (e.g., web-vue, mobile-uniapp)"/>
-    <field name="platform_type" required="true" type="string" desc="Platform type (web, mobile, desktop)"/>
-    <field name="feature_docs_path" required="true" type="string" desc="Completed feature documents base path"/>
-    <field name="features_manifest_path" required="true" type="string" desc="Path to features-{platform}.json"/>
-    <field name="module_overviews_path" required="true" type="string" desc="Parent directory containing all module overview subdirectories"/>
-    <field name="output_path" required="true" type="string" desc="Output directory for pattern documents"/>
-    <field name="language" required="true" type="string" desc="User language code"/>
-  </block>
-
-  <!-- ============================================================
-       Global Constraints
-       ============================================================ -->
-  <block type="rule" id="R1" level="forbidden" desc="Document generation constraints">
-    <field name="text">Using create_file to write pattern documents directly</field>
-    <field name="text">Full-file rewrite of pattern documents</field>
-  </block>
-
-  <block type="rule" id="R2" level="mandatory" desc="Document generation requirements">
-    <field name="text">Copy template MUST execute before filling sections for every pattern document</field>
-    <field name="text">All generated pattern documents must be in the specified language</field>
-  </block>
-
-  <!-- ============================================================
-       Global Continuous Execution Rules
-       ============================================================ -->
-  <block type="rule" id="GLOBAL-R1" level="forbidden" desc="Continuous execution constraints — NEVER violate">
-    <field name="text">DO NOT ask user "Should I continue?" or "How would you like to proceed?" during execution</field>
-    <field name="text">DO NOT offer options like "Full execution / Partial / Stop" — always execute ALL tasks to completion</field>
-    <field name="text">DO NOT suggest "Due to context window limits, let me pause" — complete current task, use checkpoint for resumption</field>
-    <field name="text">DO NOT estimate workload and suggest breaking it into phases — execute ALL items in sequence</field>
-    <field name="text">DO NOT warn about "large number of files" or "this may take a while" — proceed with generation</field>
-    <field name="text">Context window management: if approaching limit, save progress to checkpoint file and resume — do NOT ask user for guidance</field>
-  </block>
-
-  <!-- ============================================================
-       Gateway: Platform Type Check
-       ============================================================ -->
-  <block type="gateway" id="G1" mode="exclusive" desc="Check platform type for UI pattern extraction">
-    <branch test="${platform_type} NOT IN ['web', 'mobile', 'desktop']" name="Backend platform">
-      <block type="event" id="E1" action="log" level="info" desc="Log skip reason">
-        <field name="message">Skipping skill: backend platforms do not have UI patterns</field>
-      </block>
-      <block type="task" id="B1" action="analyze" desc="Return skip status">
-        <field name="status" value="skipped"/>
-        <field name="reason" value="Backend platform does not have UI patterns"/>
-        <field name="output" var="summary"/>
-      </block>
-    </branch>
-    <branch test="${platform_type} IN ['web', 'mobile', 'desktop']" name="UI platform">
-
-      <!-- ============================================================
-           Step 1: Load Features Manifest
-           ============================================================ -->
-      <sequence id="S1" name="Step 1: Load Features Manifest" status="pending" desc="Read features manifest to identify completed features">
-        <block type="task" id="B2" action="read-file" desc="Read features manifest file">
-          <field name="path" value="${features_manifest_path}"/>
-          <field name="output" var="featuresManifest"/>
-        </block>
-
-        <block type="task" id="B3" action="analyze" desc="Filter completed features">
-          <field name="input" value="${featuresManifest}"/>
-          <field name="filter" value="status === 'completed'"/>
-          <field name="collect" value="featureId, module, documentPath"/>
-          <field name="output" var="completed_features"/>
-        </block>
-      </sequence>
-
-      <!-- ============================================================
-           Gateway: Check Completed Features
-           ============================================================ -->
-      <block type="gateway" id="G2" mode="guard" desc="Check if completed features exist">
-        <branch test="${completed_features.length} > 0" name="Has completed features">
-
-          <!-- Step 1.5: Create Output Directory -->
-          <block type="task" id="B4" action="run-script" desc="Create output directory structure">
-            <field name="command">node -e "require('fs').mkdirSync('${output_path}/page-types', {recursive:true}); require('fs').mkdirSync('${output_path}/components', {recursive:true}); require('fs').mkdirSync('${output_path}/layouts', {recursive:true})"</field>
-          </block>
-
-          <!-- ============================================================
-               Step 2: Read Feature Documents
-               ============================================================ -->
-          <sequence id="S2" name="Step 2: Read Feature Documents" status="pending" desc="Extract UI information from feature documents">
-            <block type="task" id="B5" action="analyze" desc="Extract UI-related information from feature documents">
-              <field name="input" value="${completed_features}"/>
-              <field name="feature_docs_path" value="${feature_docs_path}"/>
-              <field name="extract_section" value="Interface Prototype"/>
-              <field name="extract_items" value="ASCII wireframe diagrams, layout regions"/>
-              <field name="extract_section_2" value="Page Elements Table"/>
-              <field name="extract_items_2" value="Component names, types, responsibilities, interactions"/>
-              <field name="extract_section_3" value="Business Flow Description"/>
-              <field name="extract_items_3" value="User interaction sequences, navigation patterns"/>
-              <field name="output" var="ui_extracted_data"/>
-            </block>
-          </sequence>
-
-          <!-- ============================================================
-               Step 3: Read Module Overviews
-               ============================================================ -->
-          <sequence id="S3" name="Step 3: Read Module Overviews" status="pending" desc="Gather module-level context">
-            <block type="task" id="B6" action="analyze" desc="Read module overview files for context">
-              <field name="glob_pattern" value="${module_overviews_path}/*/module-overview.md"/>
-              <field name="extract_items" value="Common page structures, Shared components, Navigation patterns"/>
-              <field name="output" var="module_context"/>
-            </block>
-          </sequence>
-
-          <!-- Checkpoint: Data Collection Complete -->
-          <block type="checkpoint" id="CP1" name="data-collection-complete" desc="Data collection complete">
-            <field name="file" value="${output_path}/.progress.json"/>
-            <field name="verify" value="${ui_extracted_data} != null AND ${module_context} != null"/>
-          </block>
-
-          <!-- ============================================================
-               Step 4: Cross-Module Clustering Analysis
-               ============================================================ -->
-          <sequence id="S4" name="Step 4: Cross-Module Clustering Analysis" status="pending" desc="Identify recurring UI patterns">
-            <block type="task" id="B7" action="analyze" desc="Perform cross-module clustering analysis">
-              <field name="input" value="${ui_extracted_data}"/>
-              <field name="module_context" value="${module_context}"/>
-              <field name="clustering_approach" value="Dynamic discovery - Agent automatically identifies and categorizes pattern types based on actual analysis results"/>
-              <field name="note" value="Templates only standardize output format, do not limit pattern types"/>
-              <field name="categories_page_types" value="Pages with similar structure and purpose (e.g., list-page, form-page, detail-page, tree-list-page, dashboard-page, wizard-page)"/>
-              <field name="categories_component_patterns" value="Reusable component combinations (e.g., search-filter-bar, data-table-pagination, modal-form, drawer-detail, tab-panel)"/>
-              <field name="categories_layout_patterns" value="Repeating structural layouts (e.g., sidebar-content, topbar-sidebar-content, full-screen)"/>
-              <field name="recognition_frequency" value="Pattern appears in 2+ features (stronger signal if across different modules)"/>
-              <field name="recognition_similarity" value="Structural similarity in ASCII wireframes"/>
-              <field name="recognition_semantic" value="Similar business purpose and interaction flow"/>
-              <field name="output" var="identified_patterns"/>
-            </block>
-          </sequence>
-
-          <!-- ============================================================
-               Gateway: Check Identified Patterns
-               ============================================================ -->
-          <block type="gateway" id="G3" mode="exclusive" desc="Check if patterns were identified">
-            <branch test="${identified_patterns.length} == 0" name="No patterns">
-              <block type="event" id="E2" action="log" level="warn" desc="Log no patterns">
-                <field name="message">No patterns identified from feature documents</field>
-              </block>
-              <block type="task" id="B8" action="analyze" desc="Return empty result">
-                <field name="status" value="completed"/>
-                <field name="platform_id" value="${platform_id}"/>
-                <field name="patterns_page_types" value='{"count": 0, "files": []}'/>
-                <field name="patterns_components" value='{"count": 0, "files": []}'/>
-                <field name="patterns_layouts" value='{"count": 0, "files": []}'/>
-                <field name="total_patterns" value="0"/>
-                <field name="output_path" value="${output_path}"/>
-                <field name="output" var="summary"/>
-              </block>
-            </branch>
-            <branch test="${identified_patterns.length} > 0" name="Has patterns">
-
-              <!-- ============================================================
-                   Step 5: Generate Pattern Documents
-                   ============================================================ -->
-              <block type="loop" id="L1" over="${identified_patterns}" as="pattern" desc="Generate pattern documents">
-
-                <!-- 5.1 Template Selection -->
-                <block type="gateway" id="G4" mode="exclusive" desc="Select template based on pattern category">
-                  <branch test="${pattern.category} == 'page-types'" name="Page type">
-                    <block type="task" id="B9" action="analyze" desc="Select page type template">
-                      <field name="template_path" value="./templates/PAGE-TYPE-TEMPLATE.md"/>
-                      <field name="output_directory" value="${output_path}/page-types/"/>
-                      <field name="output" var="selected_template"/>
-                    </block>
-                  </branch>
-                  <branch test="${pattern.category} == 'components'" name="Component pattern">
-                    <block type="task" id="B10" action="analyze" desc="Select component pattern template">
-                      <field name="template_path" value="./templates/COMPONENT-PATTERN-TEMPLATE.md"/>
-                      <field name="output_directory" value="${output_path}/components/"/>
-                      <field name="output" var="selected_template"/>
-                    </block>
-                  </branch>
-                  <branch test="${pattern.category} == 'layouts'" name="Layout pattern">
-                    <block type="task" id="B11" action="analyze" desc="Select layout pattern template">
-                      <field name="template_path" value="./templates/LAYOUT-PATTERN-TEMPLATE.md"/>
-                      <field name="output_directory" value="${output_path}/layouts/"/>
-                      <field name="output" var="selected_template"/>
-                    </block>
-                  </branch>
-                </block>
-
-                <!-- 5.2 Copy Template to Document Path -->
-                <block type="task" id="B12" action="run-script" desc="Copy template to output path">
-                  <field name="template" value="${selected_template}"/>
-                  <field name="filename_format" value="kebab-case (e.g., list-page.md, search-filter-bar.md, sidebar-content.md)"/>
-                  <field name="output_path" value="${selected_template.output_directory}/${pattern.name}.md"/>
-                  <field name="output" var="document_path"/>
-                </block>
-
-                <!-- 5.3 Fill Sections Using search_replace -->
-                <block type="task" id="B13" action="analyze" desc="Fill pattern document sections">
-                  <field name="document_path" value="${document_path}"/>
-                  <field name="pattern" value="${pattern}"/>
-                  <field name="language" value="${language}"/>
-                  <field name="constraint_forbidden" value="Using create_file to rewrite the entire document"/>
-                  <field name="constraint_mandatory" value="Use search_replace to fill each section individually"/>
-                  <field name="constraint_sections" value="All section titles MUST be preserved"/>
-                  <field name="content_ascii_wireframes" value="Must be generalized versions (not direct copies from specific features)"/>
-                  <field name="content_instance_references" value="Must use relative paths to reference actual feature documents"/>
-                  <field name="mermaid_rule" value="Follow speccrew-workspace/docs/rules/mermaid-rule.md rules"/>
-                  <field name="mermaid_syntax" value="Use graph TB/LR syntax only"/>
-                  <field name="mermaid_no_br" value="No br/ tags, no style definitions, no nested subgraph"/>
-                  <field name="mermaid_no_direction" value="No direction keyword, no special symbols"/>
-                  <field name="output" var="filled_document"/>
-                </block>
-
-                <!-- Checkpoint: Pattern Document Generated -->
-                <block type="checkpoint" id="CP2" name="pattern-document-generated" desc="Pattern document generated">
-                  <field name="file" value="${output_path}/.progress.json"/>
-                  <field name="verify" value="file_exists(${document_path}) AND document_has_all_sections(${document_path})"/>
-                </block>
-
-              </block>
-
-              <!-- Step 6: Return Summary -->
-              <block type="task" id="B14" action="analyze" desc="Generate summary of pattern documents">
-                <field name="collect_files" value="All generated file paths"/>
-                <field name="collect_counts" value="Pattern counts by category"/>
-                <field name="output" var="summary"/>
-              </block>
-
-            </branch>
-          </block>
-
-        </branch>
-        <branch test="${completed_features.length} == 0" name="No completed features">
-          <block type="event" id="E3" action="log" level="warn" desc="Log no features">
-            <field name="message">No completed features found - returning empty result</field>
-          </block>
-          <block type="task" id="B15" action="analyze" desc="Return empty summary">
-            <field name="status" value="completed"/>
-            <field name="platform_id" value="${platform_id}"/>
-            <field name="patterns_page_types" value='{"count": 0, "files": []}'/>
-            <field name="patterns_components" value='{"count": 0, "files": []}'/>
-            <field name="patterns_layouts" value='{"count": 0, "files": []}'/>
-            <field name="total_patterns" value="0"/>
-            <field name="output_path" value="${output_path}"/>
-            <field name="output" var="summary"/>
-          </block>
-        </branch>
-      </block>
-
-    </branch>
-  </block>
-
-  <!-- ============================================================
-       Output Results
-       ============================================================ -->
-  <block type="output" id="O1" desc="Workflow output results">
-    <field name="final_summary" from="${summary}" type="object" desc="Summary of generated pattern documents with file list and counts"/>
-  </block>
-
-</workflow>
-
-## Generation Rules
-
-1. **Pattern Quality**:
-   - Each pattern must have clear applicable scenarios
-   - Generalized ASCII wireframe (not feature-specific)
-   - At least 2 instance references (from same or different modules)
-
-2. **Template Compliance**:
-   - All sections from template must be filled
-   - Instance reference paths must be relative and valid
-
-3. **Mermaid Compliance**:
-   - Follow all rules in `mermaid-rule.md`
-   - Use basic `graph TB` or `graph LR` syntax
-   - No prohibited syntax elements
-
-4. **Language Consistency**:
-   - All content in the specified `language`
-   - Template section headers remain in English
-   - Content text matches user language
-
-## Error Handling
-
-| Scenario | Handling |
-|----------|----------|
-| No completed features | Return empty result, log warning |
-| No patterns identified | Return empty result, log message |
-| Template not found | Use default structure, log warning |
-| Feature document missing | Skip feature, continue with others |
-
-## Checklist
-
-- [ ] Step 1: Features manifest loaded, completed features identified
-- [ ] Step 2: All completed feature documents read
-- [ ] Step 3: All module overviews read
-- [ ] Step 4: Cross-module clustering analysis completed
-- [ ] Step 5: Pattern documents generated with correct templates
-- [ ] Step 5: File naming follows kebab-case convention
-- [ ] Step 5: ASCII wireframes are generalized versions
-- [ ] Step 5: Instance references use relative paths
-- [ ] Step 5: Mermaid diagrams follow mermaid-rule.md
-- [ ] Step 6: Summary returned with file list
-
-## Return
-
-After completion, return a summary object to the caller:
-
-```json
-{
-  "status": "completed",
-  "platform_id": "web-vue",
-  "patterns": {
-    "page_types": {
-      "count": 3,
-      "files": ["page-types/list-page.md", "page-types/form-page.md", "page-types/detail-page.md"]
-    },
-    "components": {
-      "count": 2,
-      "files": ["components/search-filter-bar.md", "components/modal-form.md"]
-    },
-    "layouts": {
-      "count": 1,
-      "files": ["layouts/sidebar-content.md"]
-    }
-  },
-  "total_patterns": 6,
-  "output_path": "speccrew-workspace/knowledges/techs/web-vue/ui-style-patterns/"
-}
-```
+> **REQUIRED**: Before executing this workflow, read the XML workflow specification: speccrew-workspace/docs/rules/agentflow-spec.md
+> Then read and execute the XML workflow in SKILL.xml block-by-block as the authoritative execution plan.

package/.speccrew/skills/speccrew-knowledge-graph-query/SKILL.md

@@ -4,107 +4,12 @@ description: Query the knowledge graph to find nodes, edges, relationships, and
 tools: Bash
 ---
 
-# Knowledge Graph Query
-
-Query the knowledge graph stored in `speccrew-workspace/knowledges/bizs/graph/` to find nodes, edges, and trace relationships. This skill wraps `graph-query.js` script for all read operations.
-
-## Trigger Scenarios
-
+# Trigger Scenarios
 - "Find all APIs in module {module}"
 - "What pages call API {apiId}?"
 - "Show impact analysis for table {tableId}"
-- "Search for entities related to {keyword}"
-- "Trace upstream dependencies of {nodeId}"
-- "Get all edges for node {nodeId}"
-
-## User
-
-Any Agent (speccrew-task-worker, speccrew-knowledge-dispatch, or other agents)
-
-## Input Variables
-
-| Variable | Type | Description | Example |
-|----------|------|-------------|---------|
-| `{{action}}` | string | Query action to perform | `"get-node"`, `"query-nodes"`, `"get-edges"`, `"search"`, `"trace-upstream"`, `"trace-downstream"` |
-| `{{id}}` | string | Node ID (for get-node, trace-*) | `"api-system-user-list"` |
-| `{{module}}` | string | Filter by module (for query-nodes, search) | `"system"`, `"trade"` |
-| `{{type}}` | string | Filter by node type (for query-nodes, search) | `"api"`, `"page"`, `"table"` |
-| `{{keyword}}` | string | Search keyword (for search) | `"user"`, `"order"` |
-| `{{direction}}` | string | Edge direction (for get-edges) | `"in"`, `"out"`, `"both"` |
-| `{{depth}}` | number | Trace depth (for trace-*) | `2`, `3` |
-| `{{graphRoot}}` | string | Path to graph root directory | `"speccrew-workspace/knowledges/bizs/graph"` |
-
-## Output Variables
-
-| Variable | Type | Description |
-|----------|------|-------------|
-| `{{status}}` | string | Query result: `"success"` or `"not-found"` |
-| `{{resultCount}}` | number | Number of results returned |
-| `{{data}}` | object/array | Query result data |
-
-## Output
-
-**Return Value (JSON format):**
-```json
-{
-  "status": "success",
-  "action": "query-nodes",
-  "resultCount": 5,
-  "data": [
-    {
-      "id": "api-system-user-list",
-      "type": "api",
-      "name": "List Users API",
-      "module": "system",
-      "sourcePath": "yudao-module-system/.../UserController.java",
-      "description": "Paginated query of system users"
-    }
-  ]
-}
-```
 
 ## AgentFlow Definition
-
 <!-- @agentflow: SKILL.xml -->
-
-> **REQUIRED**: Before executing this workflow, read the XML workflow specification: `speccrew-workspace/docs/rules/agentflow-spec.md`
-
----
-
-## Common Query Patterns
-
-| Scenario | Action | Parameters | Agent Use Case |
-|----------|--------|-----------|---------------|
-| Page → API | `get-edges` | nodeId=page-*, direction=out | UI agent: which APIs does this page use |
-| API → Table | `get-edges` | nodeId=api-*, direction=out | Dev: which tables does this API touch |
-| Table Impact | `trace-upstream` | id=table-*, depth=3 | Assess schema change impact |
-| Module Overview | `query-nodes` | module=system | Generate module summary |
-| Cross-Module Deps | `get-edges` | direction=out, filter cross-module | Inter-module dependency analysis |
-| Keyword Search | `search` | keyword=user, type=api | Find all user-related APIs |
-
----
-
-## Script Reference
-
-Scripts location: `scripts/graph-query.js` (relative to this skill directory)
-
-| Action | Parameters | Returns |
-|--------|-----------|---------|
-| `get-node` | `--id <id> --graphRoot <root>` | Single node JSON |
-| `query-nodes` | `[--module <m>] [--type <t>] --graphRoot <root>` | Node array |
-| `get-edges` | `--nodeId <id> --direction <in\|out\|both> --graphRoot <root>` | Edge array |
-| `search` | `--keyword <kw> [--type <t>] [--module <m>] --graphRoot <root>` | Node array |
-| `trace-upstream` | `--id <id> [--depth <n>] --graphRoot <root>` | Node + Edge tree |
-| `trace-downstream` | `--id <id> [--depth <n>] --graphRoot <root>` | Node + Edge tree |
-
----
-
-## Checklist
-
-- [ ] `{{action}}` is one of: `get-node`, `query-nodes`, `get-edges`, `search`, `trace-upstream`, `trace-downstream`
-- [ ] `{{graphRoot}}` path is correct (`speccrew-workspace/knowledges/bizs/graph`)
-- [ ] For `get-node` / `trace-*`: `{{id}}` follows `{type}-{module}-{name}` format
-- [ ] For `get-edges`: `{{direction}}` is one of `in`, `out`, `both`
-- [ ] For `trace-*`: `{{depth}}` is a positive integer (default: 2)
-- [ ] Script executed with Node.js (`node graph-query.js`)
-- [ ] Result JSON parsed and used by calling agent
+> **REQUIRED**: Before executing this workflow, read the XML workflow specification: speccrew-workspace/docs/rules/agentflow-spec.md
+> Then read and execute the XML workflow in SKILL.xml block-by-block as the authoritative execution plan.

package/.speccrew/skills/speccrew-knowledge-graph-write/SKILL.md

@@ -4,101 +4,12 @@ description: Write, update, and initialize knowledge graph data (nodes, edges, i
 tools: Bash
 ---
 
-# Knowledge Graph Write
-
-Write graph data (nodes and edges) to the knowledge graph storage. This skill wraps `graph-write.js` script to perform all write operations on `speccrew-workspace/knowledges/bizs/graph/`.
-
-## Trigger Scenarios
-
+# Trigger Scenarios
 - "Write graphData from skill analysis result to graph"
 - "Initialize graph structure for module {module}"
 - "Batch add nodes and edges for module {module}"
-- "Update node {nodeId} in graph"
-- "Remove node {nodeId} from graph"
-
-## User
-
-Dispatch Agent (speccrew-knowledge-dispatch)
-
-## Input Variables
-
-| Variable | Type | Description | Example |
-|----------|------|-------------|---------|
-| `{{action}}` | string | Write action to perform | `"batch-write"`, `"init-module"`, `"update-node"`, `"remove-node"` |
-| `{{platformId}}` | string | Platform identifier for directory segregation | `"backend-fastapi"`, `"backend-spring"`, `"web-vue"`, `"mobile-uniapp"` |
-| `{{module}}` | string | Target business module | `"system"`, `"trade"`, `"infra"` |
-| `{{graphData}}` | object | Graph data from skill output (for batch-write) | `{ "nodes": [...], "edges": [...] }` |
-| `{{nodeId}}` | string | Node ID (for update-node / remove-node) | `"api-system-user-list"` |
-| `{{nodeData}}` | object | Updated node data (for update-node) | `{ "description": "..." }` |
-| `{{graphRoot}}` | string | Path to graph root directory | `"speccrew-workspace/knowledges/bizs/graph"` |
-
-## Output Variables
-
-| Variable | Type | Description |
-|----------|------|-------------|
-| `{{status}}` | string | Operation result: `"success"` or `"failed"` |
-| `{{nodesWritten}}` | number | Number of nodes written |
-| `{{edgesWritten}}` | number | Number of edges written |
-| `{{message}}` | string | Summary message |
-
-## Output
-
-**Return Value (JSON format):**
-```json
-{
-  "status": "success",
-  "action": "batch-write",
-  "module": "system",
-  "nodesWritten": 5,
-  "edgesWritten": 8,
-  "message": "Successfully wrote 5 nodes and 8 edges to system module"
-}
-```
 
 ## AgentFlow Definition
-
 <!-- @agentflow: SKILL.xml -->
-
-> **REQUIRED**: Before executing this workflow, read the XML workflow specification: `speccrew-workspace/docs/rules/agentflow-spec.md`
-
----
-
-## Script Reference
-
-Scripts location: `scripts/graph-write.js` (relative to this skill directory)
-
-| Action | Parameters | Description |
-|--------|-----------|-------------|
-| `batch-write` | `--platformId <p> --module <m> --file <path> --graphRoot <root>` | Batch write nodes and edges from temp file |
-| `add-nodes` | `--platformId <p> --module <m> --file <path> --graphRoot <root>` | Batch add/replace nodes from temp file |
-| `add-edges` | `--platformId <p> --module <m> --file <path> --graphRoot <root>` | Batch add/replace edges from temp file |
-| `update-node` | `--platformId <p> --id <id> --data <json> --graphRoot <root>` | Update existing node fields |
-| `remove-node` | `--platformId <p> --id <id> --graphRoot <root>` | Remove node and connected edges |
-| `init-module` | `--platformId <p> --module <m> --graphRoot <root>` | Initialize empty module files |
-| `rebuild-index` | `--graphRoot <root>` | Rebuild global index from all module files |
-| `update-meta` | `--graphRoot <root>` | Recalculate and update graph-meta.json stats |
-
----
-
-## Cross-Module Edge Handling
-
-When edges have `source` and `target` nodes belonging to different modules:
-- The script automatically detects cross-module edges
-- Cross-module edges are stored in `{{graphRoot}}/edges/{{platformId}}/cross-module.json`
-- Module membership is determined by the node ID prefix (e.g., `api-system-*` → module `system`)
-
----
-
-## Checklist
-
-- [ ] `{{action}}` is one of: `batch-write`, `init-module`, `update-node`, `remove-node`
-- [ ] `{{platformId}}` is provided (required for all write actions)
-- [ ] `{{graphRoot}}` path is correct (`speccrew-workspace/knowledges/bizs/graph`)
-- [ ] For `batch-write`: `{{graphData}}` contains valid `nodes` and `edges` arrays
-- [ ] For `batch-write`: temp file created and cleaned up after write
-- [ ] For `init-module`: module directory files created successfully under `nodes/{{platformId}}/` and `edges/{{platformId}}/`
-- [ ] For `update-node` / `remove-node`: `{{nodeId}}` follows `{type}-{module}-{name}` format
-- [ ] Script executed with Node.js (`graph-write.js`)
-- [ ] Index and metadata updated after write operations
-- [ ] Cross-module edges correctly routed to `edges/{{platformId}}/cross-module.json`
-- [ ] Return JSON with `{{status}}`, `{{nodesWritten}}`, `{{edgesWritten}}`, `{{message}}`
+> **REQUIRED**: Before executing this workflow, read the XML workflow specification: speccrew-workspace/docs/rules/agentflow-spec.md
+> Then read and execute the XML workflow in SKILL.xml block-by-block as the authoritative execution plan.