mustard-claude 2.0.13 → 3.0.1

package/templates/_backup_20260206/enforcement.md

@@ -1,320 +0,0 @@
-# Enforcement Rules
-
-> Mandatory rules for Mustard v3.0 (stack-agnostic).
-> Pipeline enforcement via memory MCP.
-
-## Enforcement Matrix
-
-| Level | Rule | Description | Details in |
-|-------|------|-------------|------------|
-| L0 | Delegation | Main Claude does NOT implement code | This file |
-| L1 | grepai | Prefer grepai for semantic search | This file |
-| L2 | Pipeline | Pipeline mandatory for features/bugs | This file |
-| L3 | Naming | Follow naming conventions | `prompts/naming.md` |
-| L4 | Validation | Code must pass static validation | `prompts/review.md` |
-| L5 | Build | Project must compile/run | `prompts/review.md` |
-| L6 | Registry | Sync registry after creating entities | `commands/sync-registry.md` |
-
----
-
-## L0 - Mandatory Delegation
-
-### Rule L0
-
-> ⛔ Main Claude does NOT implement code. ALWAYS delegate via Task.
-
-### Self-Check
-
-Before using **Write**, **Edit**, or **Bash** (for code):
-
-```text
-Question: "Am I inside an agent (Task)?"
-
-If YES → Continue
-If NO → STOP and DELEGATE
-```
-
-### Delegation Map
-
-| Request | subagent_type | model | Prompt |
-|---------|---------------|-------|--------|
-| Bug fix | `general-purpose` | opus | `prompts/bugfix.md` |
-| New feature | `general-purpose` | opus | `prompts/orchestrator.md` |
-| Backend | `general-purpose` | opus | `prompts/backend.md` |
-| Frontend | `general-purpose` | opus | `prompts/frontend.md` |
-| Database | `general-purpose` | opus | `prompts/database.md` |
-| QA/Review | `general-purpose` | opus | `prompts/review.md` |
-| Explore | `Explore` | haiku | (native) |
-
-### Correct L0 Example
-
-```text
-User: "Add email field to Person"
-
-Main Claude:
-1. Identify: It's an entity modification (feature)
-2. Delegate: Task(subagent_type="general-purpose", model="opus", prompt="[orchestrator]...")
-3. Return: Result to user
-```
-
-### Incorrect L0 Example
-
-```text
-User: "Add email field to Person"
-
-Main Claude:
-1. Read Person.cs
-2. START EDITING ← ⛔ L0 VIOLATION
-```
-
----
-
-## L1 - Mandatory grepai
-
-### Rule L1
-
-> ⛔ **Grep and Glob are BLOCKED by the `enforce-grepai.js` hook.**
-> Use ONLY grepai for ALL code searches.
-
-### Hook L1
-
-```text
-mustard/cli/templates/hooks/enforce-grepai.js
-```
-
-- **Trigger:** Attempt to use Grep or Glob
-- **Action:** BLOCKS with permissionDecision: "deny"
-- **No exceptions**
-
-### Correct L1 Usage
-
-```javascript
-// Semantic search
-grepai_search({ query: "authentication flow" })
-
-// Call tracing
-grepai_trace_callers({ symbol: "SaveContract" })
-grepai_trace_callees({ symbol: "ValidateUser" })
-grepai_trace_graph({ symbol: "ProcessPayment", depth: 2 })
-```
-
-### BLOCKED L1 Usage
-
-```bash
-# ⛔ AUTOMATICALLY BLOCKED by hook
-grep -r "authentication" .  # ⛔ BLOCKED
-Glob("**/*.tsx")            # ⛔ BLOCKED
-```
-
-### Why grepai?
-
-| Tool | Problem |
-|------|---------|
-| Grep | Simple text search, many false positives |
-| Glob | Only finds by filename |
-| grepai | Semantic search, understands context and intent |
-
----
-
-## L2 - Mandatory Pipeline (Memory MCP)
-
-### Rule L2
-
-> Every feature/bugfix goes through the pipeline, with state persisted via memory MCP.
-
-### Hook L2
-
-```text
-mustard/cli/templates/hooks/enforce-pipeline.js
-```
-
-- **Trigger:** Edit/Write on code files
-- **Action:** Asks for confirmation (ask), Claude checks memory MCP
-- **Exceptions:** .md, .json, .yaml, .claude/, mustard/, spec/
-
-### Pipeline Phases
-
-```text
-explore → (spec approved) → implement → (validation) → completed
-```
-
-| Phase | Action | Edits Allowed |
-|-------|--------|---------------|
-| explore | Analysis, create spec | ❌ Code blocked |
-| implement | Implementation | ✅ Code allowed |
-| completed | Pipeline finished | (new pipeline needed) |
-
-### Verification via Memory MCP
-
-```javascript
-// When receiving message
-const result = await mcp__memory__search_nodes({
-  query: "pipeline phase"
-});
-
-// Check phase
-if (result.entities.length === 0) {
-  // No pipeline → free analysis, edits blocked
-}
-if (result.entities[0].observations.includes("phase: implement")) {
-  // Edits allowed
-}
-```
-
-### Pipeline Commands
-
-| Command | Phase | Action |
-|---------|-------|--------|
-| `/feature` | → explore | Creates pipeline in memory |
-| `/approve` | explore → implement | Enables edits |
-| `/validate` | implement | Checks build/type-check |
-| `/complete` | implement → done | Cleans pipeline |
-| `/resume` | (any) | Loads context |
-
-### Forbidden in L2
-
-- Edit code without active pipeline
-- Edit code in "explore" phase
-- Skip spec approval
-
----
-
-## L3 - Naming Conventions
-
-### Rule L3
-
-> Every implementation MUST follow project naming conventions.
-
-**Full details in:** [prompts/naming.md](../prompts/naming.md)
-
-### Quick Summary
-
-| Type | Pattern | Example |
-|------|---------|---------|
-| Entity | PascalCase singular | `Contract` |
-| DB Table | snake_case plural | `contracts` |
-| Endpoint | kebab-case | `/api/contracts` |
-| Hook | use + camelCase | `useContracts` |
-
----
-
-## L4 - Validation
-
-### Rule L4
-
-> Code must pass static validation (lint, type-check).
-
-**Details in:** [prompts/review.md](../prompts/review.md)
-
-Validation command depends on project stack:
-
-| Stack | Command |
-|-------|---------|
-| TypeScript | `tsc --noEmit` or `pnpm type-check` |
-| Python | `mypy` or `pyright` |
-| Go | `go vet` |
-| Rust | `cargo check` |
-
----
-
-## L5 - Build
-
-### Rule L5
-
-> Project must compile/run without errors.
-
-**Details in:** [prompts/review.md](../prompts/review.md)
-
-Build command depends on stack:
-
-| Stack | Command |
-|-------|---------|
-| .NET | `dotnet build` |
-| Node.js | `npm run build` or `pnpm build` |
-| Python | `python -m py_compile` |
-| Go | `go build` |
-| Rust | `cargo build` |
-
----
-
-## L6 - Entity Registry
-
-### Rule L6
-
-> After creating/modifying entities, update the registry.
-
-```bash
-/sync-registry
-```
-
-**Details in:** [entity-registry-spec.md](./entity-registry-spec.md)
-
----
-
-## L7-L9 - SOLID Architecture (.NET)
-
-> These rules apply to .NET projects with Entity Framework.
-> Adapt according to your project's stack.
-
-### L7 - Service Does NOT Access DbContext
-
-> Service does NOT inject DbContext directly. Uses Repository + UnitOfWork.
-
-```csharp
-// CORRECT
-public class ContractService(IContractRepository repository, IUnitOfWork unitOfWork) { }
-
-// WRONG - L7 VIOLATION
-public class ContractService(AppDbContext dbContext) { }
-```
-
-**Exception:** GraphQL Resolvers can access DbContext directly (HotChocolate pattern).
-
-### L8 - Service Only Injects Its OWN Repository
-
-> Service should only inject its own Repository. Use Services for cross-entity access.
-
-```csharp
-// CORRECT
-public class ContractService(IContractRepository repository, IPartnerService partnerService) { }
-
-// WRONG - L8 VIOLATION
-public class ContractService(IContractRepository repository, IPartnerRepository partnerRepo) { }
-```
-
-### L9 - Prefer Segregated Interfaces (ISP)
-
-> Use specific interfaces when possible for better decoupling.
-
-```csharp
-// RECOMMENDED - Specific interface
-public static Task<IResult> GetContract(Guid id, IContractQueryService service) { }
-
-// Interface composition
-public interface IContractService : IContractQueryService, IContractApprovalService { }
-```
-
----
-
-## Visual Summary
-
-```text
-L0: Delegation ──────────────────────────────────────────────────┐
-L1: grepai ───────────────────────────────────────────────────────┤ ENGINE
-L2: Pipeline ─────────────────────────────────────────────────────┤ (core/)
-                                                                  │
-L3: Naming ───────────────────────────────────────────────────────┤
-L4: Validation ───────────────────────────────────────────────────┤ CONVENTIONS
-L5: Build ────────────────────────────────────────────────────────┤ (prompts/)
-L6: Registry ─────────────────────────────────────────────────────┘
-```
-
----
-
-## See Also
-
-- [prompts/naming.md](../prompts/naming.md) - Naming conventions (L3)
-- [prompts/review.md](../prompts/review.md) - Validation and build (L4/L5)
-- [prompts/backend.md](../prompts/backend.md) - Backend architecture patterns
-- [pipeline.md](./pipeline.md) - Complete pipeline
-- [entity-registry-spec.md](./entity-registry-spec.md) - Registry spec (L6)

package/templates/_backup_20260206/mustard/approve.md

@@ -1,245 +0,0 @@
-# /approve - Approve Spec
-
-> Approves the current spec and enables the implementation phase.
-> **v2.6** - Context Reset integration
-
-## Usage
-
-```
-/approve
-/approve [comment]
-/approve --no-reset    # Skip context reset suggestion
-```
-
-## What It Does
-
-1. **Searches** for active pipeline via memory MCP
-2. **Validates** that there is a spec pending approval
-3. **Saves** exploration checkpoint to memory (auto)
-4. **Updates** phase to "implement"
-5. **Suggests** context reset for clean implementation
-6. **Enables** code edits
-
-## Implementation (Memory MCP)
-
-```javascript
-// 1. Search for active pipeline
-const result = await mcp__memory__search_nodes({
-  query: "pipeline phase explore"
-});
-
-if (!result.entities.length) {
-  return "⚠️ No active pipeline found. Use /feature or /bugfix first.";
-}
-
-const pipeline = result.entities[0];
-
-// 2. Check phase
-if (!pipeline.observations.includes("phase: explore")) {
-  return "⚠️ Pipeline is not in explore phase. Use /resume to check status.";
-}
-
-// 3. AUTO: Save exploration checkpoint
-const checkpointEntity = `Checkpoint:${pipeline.name}:explore:${Date.now()}`;
-
-await mcp__memory__create_entities({
-  entities: [{
-    name: checkpointEntity,
-    entityType: "Checkpoint",
-    observations: [
-      `pipeline: ${pipeline.name}`,
-      `phase: explore`,
-      `created: ${new Date().toISOString()}`,
-      `type: auto-approve`,
-      // Extract key insights from exploration:
-      `discovered_files: ${explorationFindings.files.join(', ')}`,
-      `patterns: ${explorationFindings.patterns.join('; ')}`,
-      `dependencies: ${explorationFindings.dependencies.join('; ')}`
-    ]
-  }]
-});
-
-await mcp__memory__create_relations({
-  relations: [{
-    from: pipeline.name,
-    to: checkpointEntity,
-    relationType: "has_checkpoint"
-  }]
-});
-
-// 4. Update to implement
-await mcp__memory__add_observations({
-  observations: [{
-    entityName: pipeline.name,
-    contents: [
-      `phase: implement`,
-      `approved: ${new Date().toISOString()}`,
-      `approved_by: user`,
-      `exploration_checkpoint: ${checkpointEntity}`
-    ]
-  }]
-});
-```
-
-### Return Message
-
-```javascript
-return `✅ Spec approved!
-
-Pipeline: ${pipeline.name}
-Phase: implement (edits enabled)
-
-## 💾 Exploration Checkpoint Saved
-
-Your exploration findings have been saved to memory:
-- Discovered files: ${explorationFindings.files.length}
-- Patterns identified: ${explorationFindings.patterns.length}
-- Dependencies mapped: ${explorationFindings.dependencies.length}
-
-Checkpoint: ${checkpointEntity}
-
----
-
-## 🧹 CONTEXT RESET RECOMMENDED
-
-The exploration phase accumulated context in this conversation.
-For optimal implementation performance, consider resetting:
-
-**Option 1: Reset now (recommended)**
-Reply with: "reset" or use \`/checkpoint --reset\`
-
-**Option 2: Continue in current context**
-Reply with: "continue"
-
----
-
-## ⚠️ CRITICAL: Delegation Required
-
-You MUST delegate implementation via Task tool. DO NOT implement in this context.
-
-### Analyze Spec Dependencies
-
-1. Read the spec file to identify layers: Backend, Frontend, Database
-2. Check if Frontend depends on NEW Backend types (sequential) or existing (parallel)
-
-### Execute Implementation
-
-**If NO dependencies between layers (parallel):**
-Call MULTIPLE Tasks in ONE message:
-- Task({ description: "⚙️ Backend {name}" })
-- Task({ description: "🎨 Frontend {name}" })
-
-**If Frontend needs NEW Backend types (sequential):**
-1. Task({ description: "⚙️ Backend {name}" }) → WAIT for completion
-2. Task({ description: "🎨 Frontend {name}" }) → After backend done
-
-### After All Tasks Complete
-1. Use /validate to verify
-2. Use /complete to finalize`;
-```
-
-## Flow
-
-```text
-/feature name
-    ↓
-EXPLORE (grepai + analysis)
-    ↓
-SPEC (create spec.md)
-    ↓
-"Spec ready. Approve?"
-    ↓
-/approve  ← YOU ARE HERE
-    │
-    ├── AUTO: Save exploration checkpoint
-    ├── SUGGEST: Context reset
-    │
-    ↓
-IMPLEMENT (clean context)
-    ↓
-/validate
-    ↓
-/complete
-```
-
-## Checks
-
-| Condition | Result |
-| --------- | ------ |
-| No active pipeline | ⚠️ Error - use /feature |
-| Pipeline in "implement" | ⚠️ Already approved |
-| Pipeline in "explore" | ✅ Approve + checkpoint |
-
-## Output
-
-### Success
-
-```text
-✅ Spec approved!
-
-Pipeline: add-email-partner
-Phase: implement (edits enabled)
-
-💾 Exploration Checkpoint Saved
-- Discovered files: 12
-- Patterns identified: 3
-- Dependencies mapped: 5
-
-🧹 CONTEXT RESET RECOMMENDED
-Reply "reset" to clear context, or "continue" to proceed.
-
-⚠️ CRITICAL: Delegation Required
-You MUST delegate via Task tool. DO NOT implement in this context.
-```
-
-### Error - No Pipeline
-
-```text
-⚠️ No active pipeline found.
-
-Use /feature <name> or /bugfix <error> to start.
-```
-
-### Error - Already Approved
-
-```text
-⚠️ Pipeline already in implement phase.
-
-Use /resume to check current status.
-Use /validate to validate implementation.
-Use /complete to finalize.
-```
-
-## Context Reset Behavior
-
-When user replies "reset" after approval:
-
-1. **Saves** final exploration summary to checkpoint
-2. **Clears** conversation context (new turn starts fresh)
-3. **User types** `/resume` to continue with clean context
-4. **Resume loads** checkpoint + spec + checklist
-
-### Why Reset?
-
-| Benefit | Description |
-| ------- | ----------- |
-| Clean context | Implementation starts without exploration noise |
-| Better focus | Agent sees only spec + patterns, not exploration history |
-| Token efficiency | More room for implementation details |
-| Reduced confusion | No old exploration paths to confuse implementation |
-
-## Notes
-
-- Only **one active pipeline** at a time
-- After approval, code edits are allowed
-- The `enforce-pipeline.js` hook checks phase via memory MCP
-- To reject spec and return to exploration, use a negative comment
-- **NEW**: Exploration checkpoint saved automatically
-- **NEW**: Context reset recommended for clean implementation
-
-## See Also
-
-- [/feature](./feature.md) - Start feature pipeline
-- [/resume](./resume.md) - Resume pipeline (loads from checkpoint)
-- [/complete](./complete.md) - Finalize pipeline
-- [/checkpoint](./checkpoint.md) - Manual checkpoint (optional)