agileflow 3.4.3 → 4.0.0-alpha.2

package/src/core/commands/rpi.md

@@ -1,633 +0,0 @@
----
-description: Orchestrate Research-Plan-Implement workflow with explicit phase transitions
-argument-hint: "[TASK=<text>] [PHASE=research|plan|implement] [SKIP_RESEARCH=true|false]"
-compact_context:
-  priority: critical
-  preserve_rules:
-    - "ACTIVE COMMAND: /agileflow:rpi - RPI workflow orchestration"
-    - "Three phases: Research → Plan → Implement"
-    - "Each phase ends with NEW CONVERSATION recommendation"
-    - "Scaling detection: trivial tasks skip research"
-    - "Context health monitoring: warn at 40% utilization"
-    - "Phase artifacts compress previous phase understanding"
-  state_fields:
-    - current_phase
-    - task_description
-    - research_artifact_path
-    - plan_artifact_path
-    - complexity_level
----
-
-# /agileflow:rpi
-
-Orchestrate the Research-Plan-Implement workflow for maximum AI agent effectiveness.
-
----
-
-## Purpose
-
-Guide complex tasks through three phases, each producing a compressed artifact for the next phase. Based on [Context Engineering for Coding Agents](../10-research/20260113-context-engineering-coding-agents.md) research.
-
-**Key insight**: LLMs are stateless. Better output requires better input context. Frequent intentional compaction at phase boundaries prevents the "dumb zone" (~40% context utilization).
-
----
-
-## STEP 0: Gather Context
-
-```bash
-node .agileflow/scripts/obtain-context.js rpi
-```
-
----
-
-<!-- COMPACT_SUMMARY_START -->
-
-## COMPACT SUMMARY - RPI WORKFLOW ACTIVE
-
-**CRITICAL**: You are running `/agileflow:rpi`. Guide user through Research → Plan → Implement phases.
-
-### The Three Phases
-
-```
-┌─────────────┐     ┌─────────────┐     ┌─────────────┐
-│  RESEARCH   │ ──► │    PLAN     │ ──► │  IMPLEMENT  │
-│             │     │             │     │             │
-│ Understand  │     │ Design      │     │ Execute     │
-│ the system  │     │ the change  │     │ the plan    │
-│             │     │             │     │             │
-│ Output:     │     │ Output:     │     │ Output:     │
-│ research.md │     │ plan.md     │     │ code + tests│
-└─────────────┘     └─────────────┘     └─────────────┘
-      │                   │                   │
-      ▼                   ▼                   ▼
-  NEW CONVERSATION    NEW CONVERSATION    DONE
-```
-
-### RULE #1: SCALING DETECTION
-
-**Assess task complexity FIRST:**
-
-| Complexity | Signs | Approach |
-|------------|-------|----------|
-| Trivial | Typo, config tweak, simple fix | Just do it |
-| Simple | Single file, clear path | Plan → Implement |
-| Moderate | Multi-file, known patterns | Plan → Implement |
-| Complex | Brownfield, unfamiliar code | Research → Plan → Implement |
-| Exploratory | Unknown territory, multi-repo | Full RPI with iteration |
-
-**Skip research when**:
-- Task is clearly defined
-- You already understand the codebase area
-- Changes are isolated to 1-2 files
-- User specifies `SKIP_RESEARCH=true`
-
-### RULE #2: PHASE TRANSITIONS
-
-**After each phase, output the compaction prompt:**
-
-```markdown
----
-
-## Phase Complete: [PHASE_NAME]
-
-**Artifact saved**: `[path/to/artifact.md]`
-
-### Recommended Next Step
-
-Start a **new conversation** with this context:
-
-> I need to [NEXT_PHASE] based on the [ARTIFACT_TYPE] at `[path]`.
->
-> [Brief summary of artifact content]
-
-**Why new conversation?** Fresh context window starts in the "smart zone" (<40% utilization). The artifact compresses all previous understanding.
-
----
-```
-
-### RULE #3: PHASE OUTPUTS
-
-**Research Phase Output** → `docs/10-research/YYYYMMDD-topic.md`:
-- Exact files and line numbers that matter
-- How the system actually works (from code, not docs)
-- Key patterns and constraints discovered
-- Risks and edge cases identified
-
-**Plan Phase Output** → `.claude/plans/[slug].md`:
-- Exact steps with file paths and line numbers
-- Actual code snippets (not pseudocode)
-- Testing strategy per change
-- Edge cases and risks
-
-**Implement Phase Output** → Code + tests:
-- Follow the plan exactly
-- Test after each change
-- Don't deviate without replanning
-
-### RULE #4: CONTEXT HEALTH WARNING
-
-If conversation is getting long, warn:
-
-```markdown
-⚠️ **Context Health Warning**
-
-This conversation has grown significantly. Signs of approaching the "dumb zone":
-- We've read many files
-- Multiple exploration cycles
-- Heavy tool output
-
-**Recommendation**: Complete current phase, save artifact, start fresh.
-```
-
-### WORKFLOW
-
-**Phase 1: Research** (when needed)
-1. Use `/agileflow:research:ask TOPIC="..."` to generate research prompt
-2. User pastes results from ChatGPT/Claude/Perplexity
-3. Use `/agileflow:research:import` to save artifact
-4. Output phase transition prompt → NEW CONVERSATION
-
-**Phase 2: Plan**
-1. Reference research artifact (if exists)
-2. Use `EnterPlanMode` for read-only exploration
-3. Design implementation approach
-4. Save plan to `.claude/plans/[slug].md`
-5. Get user approval on plan
-6. Output phase transition prompt → NEW CONVERSATION
-
-**Phase 3: Implement**
-1. Reference plan artifact
-2. Execute step-by-step
-3. Test after each change
-4. Follow plan exactly (don't deviate)
-5. Mark story complete when done
-
-### ANTI-PATTERNS
-
-❌ Skip research for complex brownfield work
-❌ Plan without understanding the system first
-❌ Implement without approved plan
-❌ Continue in same conversation after phase complete
-❌ Deviate from plan during implementation
-❌ Ignore context health warnings
-
-### DO THESE INSTEAD
-
-✅ Assess complexity before choosing phases
-✅ Complete research before planning
-✅ Get plan approval before implementing
-✅ Start fresh conversation between phases
-✅ Follow plan exactly during implementation
-✅ Monitor context health throughout
-
-### KEY FILES
-
-| File | Purpose |
-|------|---------|
-| `docs/10-research/YYYYMMDD-*.md` | Research artifacts |
-| `.claude/plans/*.md` | Plan artifacts |
-| `docs/02-practices/context-engineering-rpi.md` | RPI practice guide |
-
-### REMEMBER AFTER COMPACTION
-
-- `/agileflow:rpi` IS ACTIVE - orchestrating RPI workflow
-- Three phases: Research → Plan → Implement
-- Each phase produces compressed artifact
-- NEW CONVERSATION between phases (fresh context)
-- Assess complexity first (skip phases for simple tasks)
-- Monitor context health (warn at ~40%)
-- Follow plan exactly during implementation
-
-<!-- COMPACT_SUMMARY_END -->
-
----
-
-## Arguments
-
-| Argument | Required | Description |
-|----------|----------|-------------|
-| TASK | No | Description of what to implement |
-| PHASE | No | Start at specific phase: `research`, `plan`, or `implement` |
-| SKIP_RESEARCH | No | Set `true` to skip research phase for simpler tasks |
-
----
-
-## IMMEDIATE ACTIONS
-
-Upon invocation, execute these steps:
-
-### Step 1: Assess Task Complexity
-
-Read the TASK argument (or ask user) and assess complexity:
-
-```
-Trivial     → Just do it (no RPI needed)
-Simple      → Plan → Implement
-Moderate    → Plan → Implement
-Complex     → Research → Plan → Implement
-Exploratory → Full RPI with iteration
-```
-
-**Complexity indicators**:
-- **Trivial**: Typo fix, config change, <10 lines
-- **Simple**: Single file, clear requirements, known patterns
-- **Moderate**: 2-5 files, established patterns, some exploration
-- **Complex**: Brownfield code, unfamiliar area, many files
-- **Exploratory**: Unknown territory, needs investigation, multi-repo
-
-### Step 2: Determine Starting Phase
-
-Based on complexity assessment:
-
-```
-If PHASE argument provided → Start at that phase
-If SKIP_RESEARCH=true     → Start at Plan phase
-If Trivial                → Skip RPI, just implement
-If Simple/Moderate        → Start at Plan phase
-If Complex/Exploratory    → Start at Research phase
-```
-
-### Step 3: Execute Current Phase
-
-#### Research Phase
-
-If starting research:
-
-1. **Generate research prompt**:
-```markdown
-I'll help you research this topic. Let me generate a comprehensive prompt for external AI.
-
-Running: `/agileflow:research:ask TOPIC="[task description]"`
-```
-
-2. **Wait for user to paste results**:
-```markdown
-Copy the prompt above and paste it into ChatGPT, Claude web, or Perplexity.
-
-When you get the answer, paste the results here and I'll save them.
-```
-
-3. **Import and save research**:
-```markdown
-Running: `/agileflow:research:import TOPIC="[topic]"`
-```
-
-4. **Output phase transition**:
-```markdown
----
-
-## Phase Complete: Research ✓
-
-**Artifact saved**: `docs/10-research/[date]-[topic].md`
-
-### Key Findings Summary
-- [3-5 bullet points of key discoveries]
-
-### Recommended Next Step
-
-Start a **new conversation** with this context:
-
-> I need to plan the implementation of [task] based on the research at `docs/10-research/[date]-[topic].md`.
->
-> Key constraints discovered:
-> - [constraint 1]
-> - [constraint 2]
-
-**Why new conversation?** Fresh context window starts in the "smart zone" (<40% utilization). The research artifact compresses all understanding from exploration.
-
----
-```
-
-#### Plan Phase
-
-If starting plan:
-
-1. **Load research context** (if exists):
-```markdown
-Reading research artifact: `docs/10-research/[relevant].md`
-```
-
-2. **Enter plan mode**:
-```markdown
-Entering plan mode for read-only exploration...
-
-I'll explore the codebase to design the implementation approach.
-```
-
-3. **Explore and design** using Glob, Grep, Read tools
-
-4. **Create plan artifact**:
-```markdown
-Saving plan to: `.claude/plans/[slug].md`
-```
-
-5. **Present plan for approval**:
-```markdown
-## Implementation Plan
-
-### Step 1: [Description]
-**File**: `path/to/file.ts`
-**Change**: [What to change]
-```typescript
-// Actual code snippet
-```
-**Test**: [How to verify]
-
-### Step 2: [Description]
-...
-
----
-
-**Please review this plan.**
-
-⚠️ **Leverage reminder**: A bad plan = 100+ bad lines of code. Take time to review.
-
-Approve this plan? (YES to proceed, or provide feedback)
-```
-
-6. **Output phase transition** (after approval):
-```markdown
----
-
-## Phase Complete: Plan ✓
-
-**Artifact saved**: `.claude/plans/[slug].md`
-
-### Plan Summary
-- [N] steps to implement
-- Files affected: [list]
-- Estimated complexity: [assessment]
-
-### Recommended Next Step
-
-Start a **new conversation** with this context:
-
-> Implement the plan from `.claude/plans/[slug].md`, starting with step 1.
->
-> This plan implements [brief description]. Follow each step exactly.
-
-**Why new conversation?** Fresh context window with compressed intent. The plan contains everything needed for implementation.
-
----
-```
-
-#### Implement Phase
-
-If starting implement:
-
-1. **Load plan artifact**:
-```markdown
-Reading plan: `.claude/plans/[slug].md`
-```
-
-2. **Execute step-by-step**:
-```markdown
-## Implementing Step 1 of [N]
-
-[Execute the step as defined in plan]
-```
-
-3. **Test after each step**:
-```markdown
-Running tests to verify step 1...
-```
-
-4. **Complete implementation** with smart AskUserQuestion:
-```markdown
----
-
-## Phase Complete: Implement ✓
-
-All [N] steps executed successfully.
-
-### Summary
-- [X] files modified
-- [Y] tests added/updated
-- [Z] tests passing
-```
-
-Then present smart next steps:
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "Implementation complete: [N] steps done, [X] files modified, [Z] tests passing. What next?",
-  "header": "Next step",
-  "multiSelect": false,
-  "options": [
-    {"label": "Run full test suite (Recommended)", "description": "Verify all [Z] tests still pass after [X] file changes"},
-    {"label": "🔍 Run logic audit on [X] modified files", "description": "5 analyzers check for edge cases, race conditions, type bugs"},
-    {"label": "Create PR with /agileflow:pr", "description": "[X] files changed across [N] implementation steps"},
-    {"label": "Pause here", "description": "Implementation saved, tests can be run later"}
-  ]
-}]</parameter>
-</invoke>
-```
-
-### Step 4: Monitor Context Health
-
-Throughout execution, monitor for signs of context degradation:
-
-**Warning signs**:
-- Many files read (>10)
-- Multiple exploration cycles
-- Heavy tool output (JSON, logs)
-- Repetitive or generic responses
-
-**If detected, use smart AskUserQuestion**:
-
-```xml
-<invoke name="AskUserQuestion">
-<parameter name="questions">[{
-  "question": "⚠️ Context at ~[utilization]% ([phase] phase). [files_read] files read, [tool_calls] tool calls. Save progress?",
-  "header": "Context health",
-  "multiSelect": false,
-  "options": [
-    {"label": "Save [phase] artifact and start fresh (Recommended)", "description": "New conversation starts in smart zone (<40%). Artifact preserves all understanding."},
-    {"label": "Compact context and continue", "description": "Reduce utilization but may lose some nuance"},
-    {"label": "Continue without saving", "description": "⚠️ Quality may degrade - currently at [utilization]%"},
-    {"label": "Pause here", "description": "Progress saved to [artifact_path]"}
-  ]
-}]</parameter>
-</invoke>
-```
-
----
-
-## Scaling Guide
-
-| Task Type | Recommended Approach |
-|-----------|---------------------|
-| Typo fix | Just fix it |
-| Config change | Just do it |
-| Simple bug fix | Plan → Implement |
-| New component | Plan → Implement |
-| Feature in known area | Plan → Implement |
-| Feature in unfamiliar area | Research → Plan → Implement |
-| Complex refactoring | Research → Plan → Implement |
-| Multi-repo changes | Full RPI with iteration |
-| Architectural change | Research → Plan → Implement (with ADR) |
-
----
-
-## The Dumb Zone
-
-Context utilization affects agent performance:
-
-| Utilization | Zone | Performance |
-|-------------|------|-------------|
-| 0-40% | Smart Zone | Full reasoning capability |
-| 40-70% | Diminishing Returns | Reduced quality |
-| 70%+ | Dumb Zone | Significant degradation |
-
-**Signs you're in the Dumb Zone**:
-- Agent makes obvious mistakes
-- Responses become repetitive
-- "I understand your concern" after corrections
-- Tool calls become less accurate
-
-**Solution**: Complete current phase, save artifact, start fresh.
-
----
-
-## Leverage Math
-
-```
-Bad line of code        = 1 bad line
-Bad part of plan        = 100+ bad lines
-Bad line of research    = entire direction is hosed
-```
-
-Human review should focus on highest leverage points:
-1. Research quality (is our understanding correct?)
-2. Plan quality (are these the right steps?)
-3. Code quality (implementation details)
-
----
-
-## Related Commands
-
-| Command | Purpose |
-|---------|---------|
-| `/agileflow:research:ask` | Generate research prompts |
-| `/agileflow:research:import` | Save research artifacts |
-| `/agileflow:research:analyze` | Analyze for implementation |
-| `/agileflow:mentor` | Full orchestrated development |
-| `EnterPlanMode` | Read-only exploration mode |
-
----
-
-## Expected Output
-
-### Success - RPI Workflow
-
-```
-🔄 RPI Workflow: User Authentication Feature
-══════════════════════════════════════════════════════════════
-
-Phase 1: RESEARCH (Current)
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-Gathering context...
-✓ Read existing auth code (src/api/auth.js)
-✓ Analyzed 3 related components
-✓ Reviewed security requirements
-✓ Checked team patterns (JWT preferred)
-
-Research Summary:
-- Current: Basic session auth
-- Target: JWT with refresh tokens
-- Patterns: Use existing middleware structure
-- Risks: Token storage, refresh logic
-
-AskUserQuestion:
-  question: "Research complete: 4 findings, 2 risks identified. Ready for planning phase?"
-  options:
-    - "Start planning in new conversation (Recommended)" → "Fresh context with research artifact compressed"
-    - "Continue planning in this conversation" → "Context at ~35% - still in smart zone"
-    - "Refine research first" → "4 findings may need deeper investigation"
-    - "Pause here" → "Research saved to docs/10-research/"
-
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-Phase 2: PLAN
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-Proposed Implementation:
-1. Add JWT generation in auth.js
-2. Create refresh token middleware
-3. Update login endpoint
-4. Add token validation
-5. Write tests (unit + integration)
-
-Files to modify: 4
-New files: 2
-Estimated complexity: Medium
-
-AskUserQuestion:
-  question: "Plan ready: 5 steps, 6 files, ~medium complexity. Approve and proceed?"
-  options:
-    - "Approve - start implementing in new conversation (Recommended)" → "Fresh context with plan artifact"
-    - "Approve - implement in this conversation" → "Context at ~50% - proceed with caution"
-    - "Revise step 3 (update login endpoint)" → "Adjust plan before implementing"
-    - "Pause here" → "Plan saved to .claude/plans/"
-
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-Phase 3: IMPLEMENT
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-Implementing...
-✓ Updated src/api/auth.js
-✓ Created src/middleware/jwt.js
-✓ Created src/utils/tokens.js
-✓ Added 12 tests (all passing)
-
-🎉 Implementation complete!
-
-AskUserQuestion:
-  question: "Implementation complete: 5 steps, 4 files modified, 12 tests passing. What next?"
-  options:
-    - "Run full test suite (Recommended)" → "Verify all 12 tests pass"
-    - "🔍 Run logic audit on 4 modified files" → "5 analyzers check for edge cases"
-    - "Create PR" → "4 files changed across 5 steps"
-    - "Pause here" → "Implementation saved"
-```
-
-### Success - Phase Transition
-
-```
-🔄 RPI Phase Transition
-══════════════════════════════════════════════════════════════
-
-Current: RESEARCH → Next: PLAN
-
-⚠️ Context Health: 65% (approaching threshold)
-
-AskUserQuestion:
-  question: "Context at 65% - approaching dumb zone. Save artifact and transition?"
-  options:
-    - "Save artifact and start new conversation (Recommended)" → "Fresh context in smart zone (<40%)"
-    - "Continue in this conversation" → "⚠️ Quality may degrade past 70%"
-    - "Compact context first" → "Reduce utilization, then continue"
-    - "Pause here" → "Progress saved, resume later"
-```
-
-### Error - Wrong Phase
-
-```
-❌ Error: Cannot implement without planning
-
-Current phase: RESEARCH
-
-Complete the workflow in order:
-1. RESEARCH ✅ (current)
-2. PLAN ⏳
-3. IMPLEMENT ⏳
-
-Skip to plan: /agileflow:rpi PHASE=plan
-```
-
----
-
-## Related Documentation
-
-- [Context Engineering RPI Practice](../../docs/02-practices/context-engineering-rpi.md)
-- [Context Engineering Research](../../docs/10-research/20260113-context-engineering-coding-agents.md)