symbiote-init

Symbiote Project Init

Initialize Symbiote for the current project — scan codebase and extract DNA + intent.

MCP server is already registered globally by symbiote install. This skill only scans and extracts.

Process

1. Scan the codebase and start the server
2. Write project overview
3. Scan ALL projects for developer identity
4. Build complete DNA entries, constraints, and decisions
5. Dispatch subagent with pre-built entries
6. Print one-line summary

Step 1: Scan and Start Server

Run these as SEPARATE bash commands:

npx symbiote-cli scan

Then start the server:

npx symbiote-cli serve --no-open > /dev/null 2>&1 &

Then wait and verify:

sleep 3 && curl -s http://127.0.0.1:$(cat .brain/port)/internal/health

IMPORTANT: Do NOT combine scan and serve into one command. The scan must finish and release the DB lock before the server starts.

Step 2: Write Project Overview

Write .brain/intent/overview.md — a concise project summary that gives AI tools instant context.

Read the codebase context (CLAUDE.md, package.json, top-level structure, scan results) and write a markdown file with this exact structure:

# {Project Name}

{One sentence: what this project does and who it's for.}

## Tech Stack

- {Language/runtime} ({key detail})
- {Framework} ({purpose})
- {Database/storage} ({why chosen})
- {Other significant deps}

## Architecture

- `{top-level-dir}/` — {what it contains}
- `{top-level-dir}/` — {what it contains}

## Key Patterns

- {How data flows}
- {How modules connect}

## Entry Points

- `{file}` — {what it does}

Rules: keep it under 40 lines, no frontmatter, be specific to THIS project.

Step 3: Scan ALL Projects for Developer Identity

Read the developer's full coding identity across every project:

1. ~/.claude/CLAUDE.md — global instructions
2. Every CLAUDE.md in ~/.claude/projects/*/ — all project instructions
3. Every .md file in ~/.claude/projects/*/memory/ — all memories
4. ./CLAUDE.md — current project