Initial commit
This commit is contained in:
Zhongwei Li
256
commands/prp-core-run-all.md
Normal file
256
commands/prp-core-run-all.md
Normal file
@@ -0,0 +1,256 @@
|
||||
---
|
||||
description: Run all prp core commands in sequence from feature request to pr
|
||||
---
|
||||
|
||||
# PRP Core Complete Workflow
|
||||
|
||||
Feature: $ARGUMENTS
|
||||
|
||||
## Instructions
|
||||
|
||||
Execute the complete PRP workflow from feature idea to GitHub PR. This workflow uses a mix of subagent delegation (via Task tool) and main agent execution to optimize context usage and performance.
|
||||
|
||||
**CRITICAL**: Follow the delegation strategy exactly as specified. Some steps MUST run in subagents, others MUST run in the main conversation.
|
||||
|
||||
---
|
||||
|
||||
### Step 1: Create Branch (SUBAGENT) 🌿
|
||||
|
||||
**Delegation Strategy**: Use Task tool to delegate to subagent.
|
||||
|
||||
**Why subagent**: Branch creation is a simple, isolated task that doesn't require conversation context.
|
||||
|
||||
**Task Tool Invocation**:
|
||||
```
|
||||
description: "Create conventional git branch"
|
||||
prompt: "You must execute the slash command /prp-core:prp-core-new-branch with the argument: '$ARGUMENTS'
|
||||
|
||||
Follow ALL instructions in that command exactly:
|
||||
1. Check current branch
|
||||
2. Generate conventional branch name (feat/, fix/, chore/, etc.)
|
||||
3. Handle conflicts with version suffixes if needed
|
||||
4. Create and checkout the new branch
|
||||
5. Verify creation
|
||||
|
||||
After completion, report ONLY the branch name that was created (no additional text).
|
||||
|
||||
Example output: feat/add-user-auth"
|
||||
subagent_type: "general-purpose"
|
||||
```
|
||||
|
||||
**Wait for completion**. Capture the branch name from subagent output.
|
||||
|
||||
**Validation**: Verify branch name follows conventional format (prefix/description).
|
||||
|
||||
---
|
||||
|
||||
### Step 2: Create PRP (SUBAGENT) 📝
|
||||
|
||||
**Delegation Strategy**: Use Task tool to delegate to subagent.
|
||||
|
||||
**Why subagent**: PRP creation is research-heavy and benefits from isolated context window for codebase analysis.
|
||||
|
||||
**Task Tool Invocation**:
|
||||
```
|
||||
description: "Generate comprehensive feature PRP"
|
||||
prompt: "You must execute the slash command /prp-core:prp-core-create with the argument: '$ARGUMENTS'
|
||||
|
||||
This command will guide you through creating a comprehensive Product Requirement Prompt. Follow ALL phases:
|
||||
|
||||
Phase 1: Feature Understanding
|
||||
- Analyze the feature request thoroughly
|
||||
- Create user story format
|
||||
- Assess complexity
|
||||
|
||||
Phase 2: Codebase Intelligence Gathering
|
||||
- Use specialized agents for pattern recognition
|
||||
- Analyze project structure
|
||||
- Identify dependencies and integration points
|
||||
|
||||
Phase 3: External Research & Documentation
|
||||
- Research relevant libraries and best practices
|
||||
- Gather documentation with specific anchors
|
||||
- Identify gotchas and known issues
|
||||
|
||||
Phase 4: Deep Strategic Thinking
|
||||
- Consider edge cases and design decisions
|
||||
- Plan for extensibility and maintainability
|
||||
|
||||
Phase 5: PRP Generation
|
||||
- Output comprehensive PRP file to .claude/PRPs/features/
|
||||
|
||||
After completion, report the EXACT file path of the created PRP.
|
||||
|
||||
Example output: .claude/PRPs/features/add-user-authentication.md"
|
||||
subagent_type: "general-purpose"
|
||||
```
|
||||
|
||||
**Wait for completion**. Capture the PRP file path from subagent output.
|
||||
|
||||
**Validation**:
|
||||
1. Verify PRP file exists at reported path
|
||||
2. Verify file is not empty
|
||||
3. Verify it contains required sections (STEP-BY-STEP TASKS, VALIDATION COMMANDS)
|
||||
|
||||
---
|
||||
|
||||
### Step 3: Execute PRP (MAIN AGENT) ⚙️
|
||||
|
||||
**Delegation Strategy**: DO NOT delegate. Execute directly in main conversation using SlashCommand tool.
|
||||
|
||||
**Why main agent**:
|
||||
- PRP execution is complex and long-running (10-50+ tasks)
|
||||
- Requires full conversation context and memory
|
||||
- Needs user interaction capability for clarifications
|
||||
- Must manage state across many sequential operations
|
||||
- Benefits from conversation history and continuity
|
||||
|
||||
**Direct Execution**:
|
||||
|
||||
Use the SlashCommand tool to invoke:
|
||||
```
|
||||
/prp-core:prp-core-execute [PRP-file-path-from-step-2]
|
||||
```
|
||||
|
||||
This will:
|
||||
1. Ingest the complete PRP file
|
||||
2. Execute every task in STEP-BY-STEP TASKS sequentially
|
||||
3. Validate after each task
|
||||
4. Fix failures and re-validate before proceeding
|
||||
5. Run full validation suite on completion
|
||||
6. Move completed PRP to .claude/PRPs/features/completed/
|
||||
|
||||
**IMPORTANT**:
|
||||
- Execute this step YOURSELF in the main conversation
|
||||
- DO NOT use Task tool for this step
|
||||
- Track progress with TodoWrite
|
||||
- Don't stop until entire PRP is complete and validated
|
||||
|
||||
**Wait for execution to fully complete** before proceeding to Step 4.
|
||||
|
||||
**Validation**:
|
||||
1. All tasks completed
|
||||
2. PRP moved to completed/ directory
|
||||
3. All validation commands passed
|
||||
|
||||
---
|
||||
|
||||
### Step 4: Commit Changes (SUBAGENT) 💾
|
||||
|
||||
**Delegation Strategy**: Use Task tool to delegate to subagent.
|
||||
|
||||
**Why subagent**: Commit creation is isolated and only needs current git state.
|
||||
|
||||
**Task Tool Invocation**:
|
||||
```
|
||||
description: "Create conventional commit"
|
||||
prompt: "You must execute the slash command /prp-core:prp-core-commit
|
||||
|
||||
Follow ALL instructions in that command:
|
||||
1. Review uncommitted changes with git diff
|
||||
2. Check git status
|
||||
3. Stage all changes with git add -A
|
||||
4. Create commit with conventional format: <type>: <description>
|
||||
- Types: feat, fix, docs, style, refactor, test, chore
|
||||
- Present tense, lowercase, ≤50 chars, no period
|
||||
- NEVER mention Claude Code, Anthropic, or AI
|
||||
|
||||
After completion, report the commit hash and commit message.
|
||||
|
||||
Example output:
|
||||
Commit: a1b2c3d
|
||||
Message: feat: add user authentication system"
|
||||
subagent_type: "general-purpose"
|
||||
```
|
||||
|
||||
**Wait for completion**. Capture commit hash and message.
|
||||
|
||||
**Validation**: Verify commit exists with `git log -1 --oneline`.
|
||||
|
||||
---
|
||||
|
||||
### Step 5: Create PR (SUBAGENT) 🚀
|
||||
|
||||
**Delegation Strategy**: Use Task tool to delegate to subagent.
|
||||
|
||||
**Why subagent**: PR creation is isolated and only needs git history.
|
||||
|
||||
**Task Tool Invocation**:
|
||||
```
|
||||
description: "Push and create GitHub PR"
|
||||
prompt: "You must execute the slash command /prp-core:prp-core-pr with the title generated from the feature: '$ARGUMENTS'
|
||||
|
||||
Follow ALL instructions in that command:
|
||||
1. Check current branch
|
||||
2. Review changes with git status and git diff
|
||||
3. Push to remote with git push -u origin HEAD
|
||||
4. Create GitHub PR using gh pr create with:
|
||||
- Title: Generated from feature request
|
||||
- Body: Comprehensive description with summary, changes, type, testing, checklist
|
||||
|
||||
After completion, report the PR URL.
|
||||
|
||||
Example output: https://github.com/Wirasm/PRPs-agentic-eng/pull/15"
|
||||
subagent_type: "general-purpose"
|
||||
```
|
||||
|
||||
**Wait for completion**. Capture PR URL.
|
||||
|
||||
**Validation**: Verify PR URL is accessible.
|
||||
|
||||
---
|
||||
|
||||
## Completion Report
|
||||
|
||||
After all 5 steps complete successfully, provide a comprehensive summary:
|
||||
|
||||
```
|
||||
✅ PRP Core Workflow Complete
|
||||
|
||||
Branch: [branch-name-from-step-1]
|
||||
PRP: [prp-file-path-from-step-2] → completed/
|
||||
Commit: [commit-hash-from-step-4] - [commit-message]
|
||||
PR: [pr-url-from-step-5]
|
||||
|
||||
Summary:
|
||||
- Created conventional git branch
|
||||
- Generated comprehensive PRP with deep codebase analysis
|
||||
- Executed all [X] tasks with validation
|
||||
- Committed changes with conventional format
|
||||
- Pushed and created GitHub pull request
|
||||
|
||||
The feature is now ready for review!
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Error Handling
|
||||
|
||||
If any step fails:
|
||||
1. **STOP immediately** - do not proceed to next step
|
||||
2. Report which step failed and the error message
|
||||
3. Provide actionable guidance for fixing the issue
|
||||
4. Do NOT attempt automatic retry without user confirmation
|
||||
|
||||
---
|
||||
|
||||
## Design Rationale
|
||||
|
||||
**Subagent Delegation (Steps 1, 2, 4, 5)**:
|
||||
- Simple, isolated tasks
|
||||
- Don't need conversation context
|
||||
- Benefit from fresh context windows
|
||||
- Can run independently
|
||||
|
||||
**Main Agent Execution (Step 3)**:
|
||||
- Complex, multi-task execution
|
||||
- Requires conversation continuity
|
||||
- Needs full context and state management
|
||||
- May require user interaction
|
||||
- Long-running operation (10-50+ sequential tasks)
|
||||
|
||||
This hybrid approach optimizes:
|
||||
- **Context window usage**: Subagents for simple tasks don't pollute main context
|
||||
- **Performance**: Isolated tasks complete faster in dedicated contexts
|
||||
- **Reliability**: Complex execution stays in main agent with full capabilities
|
||||
- **User experience**: Main agent maintains conversation for complex work
|
||||
Reference in New Issue
Block a user