@avesta-hq/prevention 0.1.0

package/.claude/agents/avesta-visionary.md

@@ -0,0 +1,17 @@
+---
+name: avesta-visionary
+description: Define product vision with problem statement, users, and success metrics
+model: inherit
+tools: Read, Write, Edit
+maxTurns: 20
+---
+
+# avesta-visionary
+
+Load your system prompt and skills from the Prevention server:
+
+1. Call `avesta_get_prompt("visionary")` to receive your full instructions
+2. For each skill in the response's `skills_to_load`, call `avesta_get_skill("{skill-key}")`
+3. Follow the received instructions exactly
+
+You are operating as a specialized Prevention. Do not proceed without loading your prompt.

package/.claude/commands/avesta-acceptance-stage.md

@@ -0,0 +1,8 @@
+---
+description: "Set up acceptance stage workflow with version-based test execution"
+argument-hint: "[project name or context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="acceptance-stage")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-acceptance-test.md

@@ -0,0 +1,8 @@
+---
+description: "Write an Executable Specification (Test Case) in problem-domain language"
+argument-hint: "[behavior to specify]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="acceptance-test")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-bug-fix.md

@@ -0,0 +1,8 @@
+---
+description: "Bug fix workflow - reproduce, failing test, fix, verify"
+argument-hint: "<description of the bug to fix>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="bug-fix")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-code-review.md

@@ -0,0 +1,8 @@
+---
+description: "Review code for XP/CD best practices and Clean Architecture"
+argument-hint: "[file path or PR number]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="code-review")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-commit-stage.md

@@ -0,0 +1,8 @@
+---
+description: "Set up commit stage CI/CD pipeline with test pyramid enforcement"
+argument-hint: "[project name or context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="commit-stage")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-commit.md

@@ -0,0 +1,8 @@
+---
+description: "Create a git commit following conventional commit standards"
+argument-hint: "[optional commit description]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="commit")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-cycle.md

@@ -0,0 +1,8 @@
+---
+description: "Execute complete feature implementation (Phases 2-5) with TDD and review gates"
+argument-hint: "[feature name or continue from context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="cycle")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-dependency-review.md

@@ -0,0 +1,7 @@
+---
+description: "Review dependencies and generate gradual update plan"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="dependency-review")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-dora-init.md

@@ -0,0 +1,8 @@
+---
+description: "Initialize DORA metrics tracking for the project"
+argument-hint: "[project name or context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="dora-init")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-dora-report.md

@@ -0,0 +1,8 @@
+---
+description: "Display DORA metrics report for the project"
+argument-hint: "[time range or metric focus]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="dora-report")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-driver.md

@@ -0,0 +1,8 @@
+---
+description: "Implement Protocol Driver for acceptance tests"
+argument-hint: "[driver type: ui, api, or specific channel]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="driver")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-dsl.md

@@ -0,0 +1,8 @@
+---
+description: "Implement Domain Specific Language for acceptance tests"
+argument-hint: "[DSL method or domain to implement]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="dsl")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-enhance.md

@@ -0,0 +1,8 @@
+---
+description: "Enhancement workflow - extend existing functionality"
+argument-hint: "<description of the enhancement>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="enhance")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-green.md

@@ -0,0 +1,8 @@
+---
+description: "TDD Green Phase - Write MINIMAL code to pass the failing test"
+argument-hint: "[implementation hint]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="green")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-help.md

@@ -0,0 +1,5 @@
+---
+description: "Show Prevention status, available commands, and next steps"
+---
+Call `avesta_get_status()` and display the `display_banner` from the response verbatim in a code block.
+Then show phase-specific commands and next action from the response.

package/.claude/commands/avesta-init.md

@@ -0,0 +1,8 @@
+---
+description: "Initialize Prevention workflow tracking for a project"
+argument-hint: "[project-type: backend, frontend, fullstack]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="avesta-init")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-layer.md

@@ -0,0 +1,31 @@
+---
+description: "Implement a full layer - all tests first, then all production code"
+argument-hint: "[layer: domain | application | infrastructure | presentation] [backend | frontend]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="layer")`.
+
+If blocked: Show the `error_message` to the user and stop.
+
+If `gate_check.allowed` is true, orchestrate per-layer workers:
+
+1. **Parse arguments** from `$ARGUMENTS`:
+   - **Layer(s)**: If a specific layer is named (`domain`, `application`, `infrastructure`, `presentation`), implement only that layer. If `all` or no layer specified, implement all four in order: domain, application, infrastructure, presentation.
+   - **Project type**: If `backend` or `frontend` is specified, use it. Default: `backend`.
+
+2. **For each layer** (in the determined order), spawn a subagent:
+   - Use the `avesta-layer-worker` agent
+   - Pass a structured prompt: `"LAYER: <layer_name>\nPROJECT_TYPE: <project_type>\nFEATURE: $ARGUMENTS"`
+   - Wait for the worker to complete before starting the next layer
+   - After each worker completes, briefly report which layer finished and how many tests passed
+
+3. **After all layers complete**, show a summary:
+   ```
+   LAYER IMPLEMENTATION COMPLETE
+   - Domain: [status]
+   - Application: [status]
+   - Infrastructure: [status]
+   - Presentation: [status]
+   Next: /commit or /layer [next layer]
+   ```
+
+4. If `approve_gate` is present in the dispatch response, call `avesta_approve_gate` with that gate name after all layers complete.

package/.claude/commands/avesta-mutation-testing.md

@@ -0,0 +1,8 @@
+---
+description: "Analyze test effectiveness with mutation testing"
+argument-hint: "<branch or files to analyze>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="mutation-testing")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-plan.md

@@ -0,0 +1,8 @@
+---
+description: "Create implementation plan using Example Mapping and behavioral analysis"
+argument-hint: "<feature description, user story, or JIRA ticket>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="plan")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-red.md

@@ -0,0 +1,8 @@
+---
+description: "TDD Red Phase - Write ONE failing test"
+argument-hint: "[behavior to test]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="red")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-refactor.md

@@ -0,0 +1,8 @@
+---
+description: "TDD Refactor Phase - Improve code structure while keeping tests green"
+argument-hint: "[refactoring focus]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="refactor")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-release-stage.md

@@ -0,0 +1,8 @@
+---
+description: "Set up release stage CI/CD pipeline with contract verification and deployment"
+argument-hint: "[project name or context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="release-stage")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-scaffold.md

@@ -0,0 +1,8 @@
+---
+description: "Scaffold project structure, dependencies, and config files"
+argument-hint: "[project-type: backend, frontend, fullstack]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="scaffold")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-ship.md

@@ -0,0 +1,8 @@
+---
+description: "Merge completed work to main branch"
+argument-hint: "[optional PR title]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="ship")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-spike.md

@@ -0,0 +1,8 @@
+---
+description: "Technical exploration - disposable code for learning"
+argument-hint: "<technical question to explore>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="spike")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-tech-debt.md

@@ -0,0 +1,8 @@
+---
+description: "Tech debt workflow - approval tests first, then refactor"
+argument-hint: "<description of the tech debt to address>"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="tech-debt")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/.claude/commands/avesta-vision.md

@@ -0,0 +1,8 @@
+---
+description: "Define product vision with problem statement, users, and success metrics"
+argument-hint: "[product or feature context]"
+---
+Call `avesta_dispatch(user_input="$ARGUMENTS", current_command="vision")`.
+- If `gate_check.allowed` is true: Spawn the agent named in the `agent` field to handle this task. Pass `$ARGUMENTS` as context.
+- If blocked: Show the `error_message` to the user and stop.
+- After the agent completes successfully: If `approve_gate` is present in the response, call `avesta_approve_gate` with that gate name.

package/CLAUDE.md

@@ -0,0 +1,88 @@
+# Prevention: Continuous Delivery as a Learning System
+
+> An opinionated AI development agent that embeds XP/CD best practices to achieve Elite DORA metrics.
+> **Build the product right. Ship value fast. Learn continuously.**
+
+## Vision
+
+**Problem**: AI amplifies existing organizational patterns—both good and bad. Without engineering discipline, AI accelerates chaos.
+
+**Solution**: This agent encodes proven XP/CD practices directly into the development workflow through TDD, Clean Architecture, and comprehensive testing.
+
+**Goal**: Elite DORA metrics — on-demand deployment, <1hr lead time, 0-15% failure rate, <1hr restore time.
+
+## Target Users
+
+- **CTOs/Engineering Managers**: Engineering transformation and AI ROI
+- **Tech Leads**: Best practices at team level
+- **Individual Developers**: Features with built-in quality
+
+---
+
+## Technical Stack
+
+### Backend
+
+- **Language**: TypeScript
+- **Runtime**: Node.js
+- **Framework**: Express
+- **Testing**: Jest (unit/component/contract), Supertest
+- **Contract Testing**: Pact (Provider)
+
+### Frontend
+
+- **Language**: TypeScript
+- **Framework**: Next.js (App Router)
+- **UI Library**: shadcn/ui + Radix UI
+- **Styling**: Tailwind CSS
+- **Testing**: Jest + React Testing Library
+- **Contract Testing**: Pact (Consumer)
+
+### Shared
+
+- **Package Manager**: pnpm
+- **Linting**: ESLint + Prettier
+- **Acceptance Testing**: Cucumber + Playwright
+- **External System Mocking**: [Scenarist](https://scenarist.io/)
+
+---
+
+## Rules for Claude
+
+### General
+
+- Use pnpm for package management
+- Follow conventional commits (feat:, fix:, refactor:, test:, docs:)
+- Never mention TDD in code comments or commit messages—the process is invisible
+- All code must have tests at the appropriate pyramid level
+- **NO "WHAT" comments**: Only add "WHY" comments when business logic or technical decisions are non-obvious
+
+### File Naming
+
+- Use kebab-case for files: `add-project.use-case.ts`
+- Tests co-located with implementation: `*.test.ts`
+- Use `.types.ts` for type definitions
+- Use `.factory.ts` for composition root (in `presentation/factories/`, NOT in `application/`)
+
+---
+
+## MCP Server Integration
+
+This project uses Prevention as an MCP server with 12 tools, 27 agents, and 27 skills that provide all methodology knowledge (TDD, Clean Architecture, test pyramid, CI/CD, etc.) on demand.
+
+**When starting work:**
+
+1. Run `/avesta-help` to see current status and next steps
+2. The workflow tracks your phase (vision → plan → atdd → tdd → review → ship)
+3. Gates enforce that prerequisites are met before advancing
+
+**Key tools:**
+
+- `avesta_get_status` — Current phase, gates, next action
+- `avesta_dispatch` — Route tasks to the right workflow step
+- `avesta_get_catalog` — All available commands and skills
+- `avesta_get_skill` — Deep knowledge on any practice (testing, architecture, CI/CD)
+
+**All detailed methodology is in skills** — use `avesta_get_skill` or `avesta_get_catalog` to access TDD rules, Clean Architecture patterns, test strategies, CI/CD pipeline templates, etc.
+
+

package/README.md

@@ -0,0 +1,113 @@
+# Prevention
+
+> **AI development agent that embeds XP/CD best practices to achieve Elite DORA metrics.**
+
+Build the product right. Ship value fast. Learn continuously.
+
+---
+
+## What is Prevention?
+
+Prevention turns [Claude Code](https://claude.ai/claude-code) into a disciplined engineering partner. Instead of just generating code, it enforces **TDD, Clean Architecture, ATDD, and Continuous Delivery** — the practices that separate elite teams from the rest.
+
+One command to set up. Slash commands to drive the workflow. Gates to keep you honest.
+
+---
+
+## Quick Start
+
+```bash
+cd your-project
+npx @avesta-hq/prevention init
+```
+
+Then start Claude Code and run `/avesta-help` to see your status and available commands.
+
+---
+
+## Workflow
+
+Prevention guides you through a proven delivery workflow:
+
+```
+vision → plan → acceptance tests → TDD → driver → review → ship
+```
+
+Each phase has gates that ensure prerequisites are met before you move forward. You can't skip writing tests before shipping, or ship without a review.
+
+---
+
+## Commands
+
+27 slash commands organized by workflow phase:
+
+| Phase                | Command                           | Purpose                                              |
+| -------------------- | --------------------------------- | ---------------------------------------------------- |
+| **Setup**            | `/avesta-init [type]`             | Initialize workflow tracking                         |
+|                      | `/avesta-scaffold`                | Scaffold project structure (greenfield)              |
+|                      | `/avesta-help`                    | Show status, gates, and next steps                   |
+| **Vision**           | `/avesta-vision`                  | Define product vision, success metrics, scope        |
+| **Planning**         | `/avesta-plan <feature>`          | Break feature into TDD-ready tasks (Example Mapping) |
+|                      | `/avesta-spike <question>`        | Technical exploration (disposable code)              |
+| **Acceptance Tests** | `/avesta-acceptance-test`         | Write Executable Specification                       |
+|                      | `/avesta-dsl`                     | Implement Domain Specific Language layer             |
+|                      | `/avesta-driver`                  | Implement Protocol Drivers (UI/API)                  |
+| **TDD**              | `/avesta-red <behavior>`          | Write ONE failing test                               |
+|                      | `/avesta-green`                   | Minimal code to make it pass                         |
+|                      | `/avesta-refactor`                | Improve structure (tests must stay green)            |
+|                      | `/avesta-cycle <behavior>`        | Full red-green-refactor cycle                        |
+|                      | `/avesta-layer`                   | All tests first, then all production code            |
+| **Change Types**     | `/avesta-bug-fix <description>`   | Bug fix workflow                                     |
+|                      | `/avesta-enhance <description>`   | Enhancement workflow                                 |
+|                      | `/avesta-tech-debt <description>` | Tech debt workflow (approval tests)                  |
+| **Review & Ship**    | `/avesta-code-review`             | Domain-specific code review                          |
+|                      | `/avesta-commit`                  | Conventional commit                                  |
+|                      | `/avesta-ship`                    | Merge to main                                        |
+| **CI/CD**            | `/avesta-commit-stage`            | Set up commit stage pipeline                         |
+|                      | `/avesta-release-stage`           | Set up release stage pipeline                        |
+|                      | `/avesta-acceptance-stage`        | Set up acceptance stage pipeline                     |
+| **Metrics**          | `/avesta-dora-init`               | Initialize DORA metrics tracking                     |
+|                      | `/avesta-dora-report`             | Generate DORA metrics report                         |
+| **Maintenance**      | `/avesta-dependency-review`       | Review deps and generate update plan                 |
+|                      | `/avesta-mutation-testing`        | Analyze test effectiveness                           |
+
+---
+
+## License Tiers
+
+| Feature              | Free                                               | Pro                  |
+| -------------------- | -------------------------------------------------- | -------------------- |
+| **Core TDD**         | `/avesta-red`, `/avesta-green`, `/avesta-refactor` | All 27 commands      |
+| **Setup**            | `/avesta-init`, `/avesta-scaffold`, `/avesta-help` | All 28 agents        |
+| **Skills**           | -                                                  | 27 contextual skills |
+| **Gate enforcement** | -                                                  | Automated gates      |
+| **Projects**         | 1                                                  | Unlimited            |
+
+Free tier gives you the core TDD loop. Pro unlocks the full workflow with all commands, skills, and gate enforcement.
+
+---
+
+## Manual MCP Setup
+
+If you prefer to configure the MCP server manually, create `.mcp.json` at your project root:
+
+```json
+{
+  "mcpServers": {
+    "prevention": {
+      "command": "npx",
+      "args": ["@avesta-hq/prevention", "serve"]
+    }
+  }
+}
+```
+
+Then restart Claude Code.
+
+---
+
+## Requirements
+
+- Node.js >= 18
+- [Claude Code](https://claude.ai/claude-code) CLI
+