# PRPs-agentic-eng **Repository Path**: mirrors_trending/PRPs-agentic-eng ## Basic Information - **Project Name**: PRPs-agentic-eng - **Description**: Prompts, workflows and more for agentic engineering - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: development - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2025-07-20 - **Last Updated**: 2026-07-25 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # PRP (Product Requirement Prompts) A collection of prompts for AI-assisted development with Claude Code. ## Video Walkthrough https://www.youtube.com/watch?v=KVOZ9s1S9Gk&lc=UgzfwxvFjo6pKEyPo1R4AaABAg ### Support This Work **Found value in these resources?** **Buy me a coffee:** https://coff.ee/wirasm I spent a considerable amount of time creating these resources and prompts. If you find value in this project, please consider buying me a coffee to support my work. --- ### Transform Your Team with AI Engineering Workshops **Ready to move beyond toy demos to production-ready AI systems?** **Book a workshop:** https://www.rasmuswiding.com/ **What you'll get:** - Put your team on a path to become AI power users - Learn the exact PRP methodology used by top engineering teams - Hands-on training with Claude Code, PRPs, and real codebases - From beginner to advanced AI engineering workshops for teams and individuals **Perfect for:** Engineering teams, Product teams, and developers who want AI that actually works in production Contact me directly at hello@rasmuswiding.com --- ## What is PRP? **Product Requirement Prompt (PRP)** = PRD + curated codebase intelligence + agent/runbook The minimum viable packet an AI needs to ship production-ready code on the first pass. A PRP supplies an AI coding agent with everything it needs to deliver a vertical slice of working software—no more, no less. ### How PRP Differs from Traditional PRD A traditional PRD clarifies _what_ the product must do and _why_ customers need it, but deliberately avoids _how_ it will be built. A PRP keeps the goal and justification sections of a PRD yet adds AI-critical layers: - **Context**: Precise file paths, library versions, code snippet examples - **Patterns**: Existing codebase conventions to follow - **Validation**: Executable commands the AI can run to verify its work --- ## Quick Start ### Option 1: Install the plugin (recommended) ```bash /plugin marketplace add Wirasm/PRPs-agentic-eng /plugin install prp-core@prp-marketplace ``` Every skill is then available across all your projects as `/prp-core:`, updated centrally. ### Option 2: Copy the skills you want ```bash # From your project root — copy individual skills (or the whole folder) cp -r /path/to/PRPs-agentic-eng/.claude/skills/prp-plan .claude/skills/ ``` ### Option 3: Clone the repository ```bash git clone https://github.com/Wirasm/PRPs-agentic-eng.git cd PRPs-agentic-eng ``` --- ## Skills The `.claude/skills/` directory contains the core PRP workflow as Agent Skills — also distributed as the `prp-core` plugin (invoke as `/prp-core:`): ### Core Workflow | Command | Description | | ---------------- | -------------------------------------------------------- | | `/prp-prd` | Interactive PRD generator with implementation phases | | `/prp-plan` | Create implementation plan (from PRD or free-form input) | | `/prp-implement` | Execute a plan with validation loops | ### Issue & Debug Workflow | Command | Description | | ------------------------ | ------------------------------------------------ | | `/prp-issue` | Investigate a GitHub issue, then implement the fix (`investigate` / `fix` verbs) | | `/prp-debug` | Deep root cause analysis with 5 Whys methodology | ### Git & Review | Command | Description | | ------------- | ------------------------------------------------- | | `/prp-commit` | Smart commit with natural language file targeting | | `/prp-pr` | Create PR with template support | | `/prp-review` | Comprehensive PR code review | ### Autonomous Loop | Command | Description | | ----------- | --------------------------------------------------------------------- | | `/prp-loop` | Autonomous cyclic pipeline: plan → implement → pr → review (review→fix loops until clean) | | `/prp-orchestrate` | Coordinate parallel workstreams running PRP skills in git worktrees, with review gates and merge sequencing | ### Research & Authoring | Command | Description | | ------------------------ | ------------------------------------------------------------------------ | | `/prp-codebase-question` | Research how the codebase works using parallel agents | | `/prp-research-team` | Design a multi-agent research team and executable research plan | | `/prp-meta-skill` | Author new skills, or refactor fat skills into lean SKILL.md + references | --- ## PRP Loop (Autonomous Execution) `/prp-loop` drives the full pipeline (`plan → implement → pr → review`) headlessly, running one `claude -p` session per stage and looping `review → fix` until the PR review comes back clean. Progress is tracked in the project's PRP store at `~/.prp//state/prp-loop.state.json`. ### How It Works ``` /prp-loop "add user authentication with JWT" ``` 1. **plan** — writes `~/.prp//plans/.plan.md` 2. **implement** — executes the plan, looping until all validations pass, then commits 3. **pr** — pushes the branch and opens the PR 4. **review** — reviews the PR and writes a `{clean, blocking}` verdict 5. **cycle** — if not clean, blocking findings feed back into a fix pass → push → re-review, until clean or `--max-cycles` is reached ### Usage ```bash # Run the full pipeline from a feature description /prp-loop "add user authentication with JWT" # Grind a single plan to green and stop before PR/review (the old Ralph behaviour) /prp-loop "add user authentication with JWT" --until implement # Resume a halted or in-progress loop /prp-loop --resume ``` ### Tips - `--until ` (`plan` | `implement` | `pr` | `review` | `fix`) halts once that stage completes; `--until implement` is the headless replacement for the retired single-session Ralph loop ("grind one plan to green, no PR") - Defaults: `--max-cycles 3`, `--max-implement-iterations 10`; base branch is auto-detected - Pass `--validate ""` to give the loop an authoritative green check (exit 0 = pass) - Works best with plans that have clear, testable validation commands - State is tracked in `~/.prp//state/prp-loop.state.json`; inspect or resume from there --- ## Workflow Overview ### Large Features: PRD → Plan → Implement ``` /prp-prd "user authentication system" ↓ Creates PRD with Implementation Phases table ↓ /prp-plan ~/.prp//prds/user-auth.prd.md ↓ Auto-selects next pending phase, creates plan ↓ /prp-implement ~/.prp//plans/user-auth-phase-1.plan.md ↓ Executes plan, updates PRD progress, archives plan ↓ Repeat /prp-plan for next phase ``` ### Medium Features: Direct to Plan ``` /prp-plan "add pagination to the API" ↓ Creates implementation plan from description ↓ /prp-implement ~/.prp//plans/add-pagination.plan.md ``` ### Bug Fixes: Issue Workflow ``` /prp-issue investigate 123 ↓ Analyzes issue, creates investigation artifact ↓ /prp-issue fix 123 ↓ Implements fix, creates PR ``` --- ## Artifacts Structure Artifacts and runtime state are stored outside the repository in a per-project directory: ``` ~/.prp// ├── project.json # Canonical project path and name ├── prds/ # Product requirement documents ├── plans/ # Implementation plans │ └── completed/ # Archived completed plans ├── research/ # Codebase research ├── research-plans/ # Multi-agent research plans ├── reports/ # Implementation reports ├── reviews/ # Human-readable PR reviews ├── issues/ # Issue investigation artifacts │ └── completed/ # Archived completed investigations ├── debug/ # Root-cause analysis reports ├── orchestration/ # Parallel-workstream run files └── state/ # Loop state, verdicts, logs, and hook sentinels ``` The project key is derived from the canonical main-checkout path, so all linked worktrees share the same store. Set `PRP_HOME` to override the default `~/.prp` root. --- ## PRD Phases PRDs include an Implementation Phases table for tracking progress: ```markdown | # | Phase | Description | Status | Parallel | Depends | PRP Plan | | --- | ----- | ----------- | ----------- | -------- | ------- | -------- | | 1 | Auth | User login | complete | - | - | [link] | | 2 | API | Endpoints | in-progress | - | 1 | [link] | | 3 | UI | Frontend | pending | with 4 | 2 | - | | 4 | Tests | Test suite | pending | with 3 | 2 | - | ``` - **Status**: `pending` → `in-progress` → `complete` - **Parallel**: Phases that can run concurrently (in separate worktrees) - **Depends**: Phases that must complete first --- ## PRP Best Practices 1. **Context is King**: Include ALL necessary documentation, examples, and caveats 2. **Validation Loops**: Provide executable tests/lints the AI can run and fix 3. **Information Dense**: Use keywords and patterns from the codebase 4. **Bounded Scope**: Each plan should be completable by an AI in one loop --- ## Project Structure ``` your-project/ ├── .claude/ │ ├── skills/ # PRP skills (or install the prp-core plugin) │ └── agents/ # Custom subagents ├── CLAUDE.md # Project-specific guidelines └── src/ # Your source code ``` --- ## Parallel Development with Worktrees When PRD phases can run in parallel: ```bash # Phase 3 and 4 can run concurrently git worktree add -b phase-3-ui ../project-phase-3 git worktree add -b phase-4-tests ../project-phase-4 # Run Claude in each cd ../project-phase-3 && claude cd ../project-phase-4 && claude ``` --- ## Resources ### Legacy Commands, Templates & Docs The previous slash-command generation is preserved in `old-prp-commands/` for reference only. That includes its PRP templates (`old-prp-commands/PRPs/templates/` — `prp_base.md`, `prp_story_task.md`, `prp_planning.md`) and curated AI documentation (`old-prp-commands/PRPs/ai_docs/`). The current skills carry their templates inline or in their own `templates/` directories. --- ## License MIT — see [LICENSE](LICENSE). --- ## Support I spent a considerable amount of time creating these resources and prompts. If you find value in this project, please consider buying me a coffee to support my work. **Buy me a coffee:** https://coff.ee/wirasm --- **The goal is one-pass implementation success through comprehensive context.**