# Project Research Skill
Structured project kickoff for Claude Code. Conducts one-question-at-a-time interviews, auto-researches tech docs via Context7 and web search, establishes AI coding guardrails, and compiles **CLAUDE.md** + **PLAN.md** before a single line of code is written.
## What it does
1. **9-phase interview** covering vision, tech stack, data model, user flows, coding style, design language, guardrails, and gap analysis
2. **Auto-researches** every technology via Context7 docs and web search (latest docs, best practices, common pitfalls)
3. **Proposes answers** for every question to reduce mental burden (multiple choice with multiSelect)
4. **Saves incrementally** after every answer to `.draft` files (crash-safe)
5. **Produces two files:**
- `CLAUDE.md` — lean <80 line cheat sheet loaded every session (only actionable rules)
- `PLAN.md` — comprehensive project document with research, data model, user flows, and implementation phases
## Install
### Option A: Plugin marketplace (recommended)
```
/plugin marketplace add agamm/project-research-skill
/plugin install project-research@project-research-marketplace
```
### Option B: Copy to personal skills
```bash
mkdir -p ~/.claude/skills/project-research
curl -sL https://raw.githubusercontent.com/agamm/project-research-skill/main/skills/project-research/SKILL.md -o ~/.claude/skills/project-research/SKILL.md
curl -sL https://raw.githubusercontent.com/agamm/project-research-skill/main/skills/project-research/interview-questions.md -o ~/.claude/skills/project-research/interview-questions.md
curl -sL https://raw.githubusercontent.com/agamm/project-research-skill/main/skills/project-research/output-templates.md -o ~/.claude/skills/project-research/output-templates.md
```
### Option C: Clone
```bash
git clone https://github.com/agamm/project-research-skill.git
cp -r project-research-skill/skills/project-research ~/.claude/skills/
```
## Usage
Start a new Claude Code session and say:
```
/project-research
```
Or just describe your project idea — Claude will detect the skill and offer to kick off the research process.
## Interview Phases
| Phase | What it covers |
|-------|---------------|
| 1. Project Vision | What, why, who, scope, platform, existing code audit |
| 2. Tech Stack + Research | Technologies, auto-research docs/best practices, compatibility check |
| 3. Data Model | Entities, relationships, data flow, lifecycle |
| 4. User Flows | Core journeys, error/empty/loading states, edge cases |
| 5. Coding Style | Paradigm, naming, testing, error handling, file organization |
| 6. Design Language | Aesthetic, colors, typography, spacing, animations (skipped for CLI/API) |
| 7. Guardrails | File limits, architecture boundaries, dependency governance, anti-patterns |
| 8. Holistic Review | AI-driven gap analysis for non-obvious concerns |
| 9. Compile | Quality-gated CLAUDE.md + PLAN.md generation |
## Key features
- **Crash protection** — saves to `PLAN.md.draft` and `CLAUDE.md.draft` after every answer. Resumable.
- **Dynamic questions** — options adapt to your project context, not hardcoded
- **Post-research reconciliation** — flags technology conflicts and suggests better alternatives
- **Quality gate** — final files are zero-duplicate, zero-fluff, scannable, and specific
## Files
```
skills/project-research/
├── SKILL.md # Main skill instructions (9 phases, interview rules)
├── interview-questions.md # Question bank with AskUserQuestion examples
└── output-templates.md # CLAUDE.md and PLAN.md templates + quality standards
```
## License
MIT