Test MCP_Perso

80_bmad/articulation_avec_lead_tech.md

@@ -173,3 +173,15 @@ $LEADTECH/95_a_capitaliser.md
   ↓ (validation humaine)
 Lead_tech (fichier approprié)
 ```
+
+---
+
+# Extension optionnelle : MCP sidecar
+
+Pour rendre la base interactive dans BMAD sans changer l'orchestration BMAD,
+un serveur MCP sidecar peut etre ajoute.
+
+Reference d'implementation dans ce repo :
+
+- `80_bmad/integration_mcp_sidecar.md`
+- `mcp/leadtech_bmad_mcp/`

80_bmad/base/_bmad/bmm/agents/dev.md

@@ -50,6 +50,7 @@ You must fully embody this agent's persona and follow all activation instruction
       <r> Stay in character until exit selected</r>
       <r> Display Menu items as the item dictates and in the order given.</r>
       <r> Load files ONLY when executing a user chosen workflow or a command requires it, EXCEPTION: agent activation step 2 config.yaml</r>
+      <r> If leadtech-bmad MCP server is available: call validate_plan before implementation and validate_patch before marking story for review. Treat blocking_issues as hard blockers — do NOT proceed past them without explicit user override.</r>
     </rules>
 </activation>  <persona>
     <role>Senior Software Engineer</role>

80_bmad/base/_bmad/bmm/agents/sm.md

@@ -49,6 +49,7 @@ You must fully embody this agent's persona and follow all activation instruction
       <r> Stay in character until exit selected</r>
       <r> Display Menu items as the item dictates and in the order given.</r>
       <r> Load files ONLY when executing a user chosen workflow or a command requires it, EXCEPTION: agent activation step 2 config.yaml</r>
+      <r> If leadtech-bmad MCP server is available: during create-story workflow, call get_guidance to enrich story sections "Patterns a appliquer", "Risques a eviter", and "Leadtech MCP Gates". If blocking_issues is non-empty, do NOT set story to ready-for-dev without explicit user acknowledgement.</r>
     </rules>
 </activation>  <persona>
     <role>Technical Scrum Master + Story Preparation Specialist</role>

80_bmad/base/_bmad/bmm/workflows/4-implementation/create-story/instructions.xml

@@ -274,6 +274,25 @@
     <template-output
       file="{default_output_file}">story_requirements</template-output>
 
+    <!-- Lead_tech MCP guidance: enrich story with validated patterns and risks -->
+    <check if="leadtech-bmad MCP server is available">
+      <action>Extract domain from story (backend|frontend|ux|n8n|product|workflow)</action>
+      <action>Call MCP tool: get_guidance(domain={{domain}}, task_type="analysis", story_text={{story_text}})</action>
+
+      <!-- Lecture des docs complets pour les top matches -->
+      <action>For each URI listed in the "Lire le contenu complet" should_do of get_guidance response:
+        - If URI starts with "leadtech://knowledge/": call MCP resource leadtech://knowledge/{domain}/{bucket}/{slug}
+        - If URI starts with "leadtech://global/": call MCP resource leadtech://global/{label}
+        Read the full content and use it to enrich the story dev notes and guardrails below.
+      </action>
+
+      <action>Write result into story section "Patterns a appliquer": must_do items from get_guidance response, including excerpts</action>
+      <action>Write result into story section "Risques a eviter": red_flags items from get_guidance response, including excerpts</action>
+      <action>If global docs matched (should_do contains "Consulter doc global"): add a "Contexte Lead_tech global" section in Dev Notes with the relevant excerpts</action>
+      <action>Write result into story section "Leadtech MCP Gates": { tool: "get_guidance", timestamp: now, must_do, red_flags, blocking_issues, decision: "voir implementation" }</action>
+      <critical>Si blocking_issues est non vide, signaler au SM avant de passer ready-for-dev</critical>
+    </check>
+
     <!-- Developer context section - MOST IMPORTANT PART -->
     <template-output file="{default_output_file}">
     developer_context_section</template-output> **DEV AGENT GUARDRAILS:** <template-output file="{default_output_file}">

80_bmad/base/_bmad/bmm/workflows/4-implementation/dev-story/instructions.xml

@@ -227,6 +227,24 @@
     <!-- Story parallelization check -->
     <action>Before implementation, read `Parallel-safe`, `Depends-on`, and `Can-run-with` from the story file.</action>
     <action>If metadata conflicts with the actual code state or branch reality, stop and reconcile the story before coding.</action>
+
+    <!-- Lead_tech MCP gate: validate plan before implementation -->
+    <check if="leadtech-bmad MCP server is available">
+      <action>Extract domain from story file (backend|frontend|ux|n8n|product|workflow)</action>
+      <action>Build plan_text: concatenate story tasks/subtasks and Dev Notes sections</action>
+      <action>Call MCP tool: validate_plan(domain={{domain}}, plan_text={{plan_text}}, agent_role="builder", strict=true)</action>
+      <action>Write gate result into story section "Leadtech MCP Gates": { tool: "validate_plan", timestamp: now, must_do, red_flags, blocking_issues }</action>
+      <critical>Si blocking_issues est non vide: HALT — corriger le plan avant de commencer l'implémentation</critical>
+
+      <!-- Lecture des docs référencés dans must_do/red_flags si URIs MCP présentes -->
+      <action>Scan must_do and red_flags for URIs of the form "leadtech://knowledge/..." or "leadtech://global/...".
+        For each URI found: read the full document via the corresponding MCP resource.
+        Use the content to reinforce implementation constraints — treat it as authoritative Lead_tech guidance.
+      </action>
+
+      <action>Appliquer chaque must_do du résultat comme contrainte d'implémentation</action>
+    </check>
+
   <step n="5" goal="Implement task following red-green-refactor cycle">
     <critical>FOLLOW THE STORY FILE TASKS/SUBTASKS SEQUENCE EXACTLY AS WRITTEN - NO DEVIATION</critical>
 
@@ -330,6 +348,17 @@
     <action>Verify ALL tasks and subtasks are marked [x] (re-scan the story document now)</action>
     <action>Run the full regression suite (do not skip)</action>
     <action>Confirm File List includes every changed file</action>
+
+    <!-- Lead_tech MCP gate: validate patch before marking for review -->
+    <check if="leadtech-bmad MCP server is available">
+      <action>Produce diff_text: concatenate git diff or list of changes made during implementation</action>
+      <action>Extract changed_files list from story File List section</action>
+      <action>Call MCP tool: validate_patch(domain={{domain}}, diff_text={{diff_text}}, changed_files={{changed_files}}, strict=true)</action>
+      <action>Write gate result into story section "Leadtech MCP Gates": { tool: "validate_patch", timestamp: now, must_do, red_flags, blocking_issues }</action>
+      <critical>Si blocking_issues est non vide: HALT — corriger avant de passer en review</critical>
+      <action>Si must_do non vide et non déjà couverts: traiter comme tâches complémentaires avant review</action>
+    </check>
+
     <action>Execute enhanced definition-of-done validation</action>
     <action>Update the story Status to: "review"</action>
 

80_bmad/integration_mcp_sidecar.md

@@ -0,0 +1,57 @@
+# BMAD — Integration MCP Sidecar (Lead_tech)
+
+Ce document decrit le cablage du serveur MCP `leadtech-bmad-mcp` dans BMAD.
+
+## Positionnement
+
+- BMAD reste le systeme d'orchestration.
+- Le MCP Lead_tech est un **appui**: guidance, controles qualite, capitalisation.
+- Le MCP ne pilote ni branche, ni statut, ni execution des stories.
+
+## Points d'injection
+
+1. Entrée story (Analyst)
+- Appel: `get_guidance(domain, task_type="analysis", story_text=...)`
+- Sortie injectee dans la story:
+  - `Patterns a appliquer`
+  - `Risques a eviter`
+  - `Gates de validation`
+
+2. Pre-implementation (Builder)
+- Appel: `validate_plan(domain, plan_text, agent_role="builder")`
+- Regle: pas d'implementation tant que `blocking_issues` n'est pas vide.
+
+3. Post-patch (Builder)
+- Appel: `validate_patch(domain, diff_text, changed_files)`
+- Resultat ajoute au Dev Agent Record.
+
+4. Review (Reviewer)
+- Appel: `emit_checklist(agent_role="reviewer", domain, story_text)`
+- Option: relancer `validate_patch` sur le diff final de PR.
+
+5. Cloture et apprentissage (Curator)
+- Appel: `propose_capitalization(...)` en `dry_run=true` par defaut.
+- Revue periodique: `triage_capitalization()`.
+
+## Mode de rollout recommande
+
+- Phase 1: advisory only (aucun blocage automatique)
+- Phase 2: blocage sur `blocking_issues` uniquement
+- Phase 3: rules strictes sur domaines critiques (backend auth/contracts/sessions)
+
+## Traceabilite
+
+Chaque story BMAD doit contenir une section `Leadtech MCP Gates` avec:
+
+- timestamp
+- tool appele
+- resume des `must_do`, `red_flags`, `blocking_issues`
+- decision humaine associee
+
+## Contraintes de securite
+
+- `LEADTECH_MCP_ALLOW_WRITE=0` en environnement normal.
+- Ecriture activee ponctuellement uniquement pour:
+  - `propose_capitalization(dry_run=false)`
+  - `route_to_project_memory(dry_run=false)`
+- Jamais d'ecriture directe dans `knowledge/*` par tool MCP.