optimizing-code

```bash

cd your-project

curl -sSL https://raw.githubusercontent.com/dralgorhythm/claude-agentic-framework/main/scripts/init-framework.sh | bash -s .

```

Then start Claude Code: claude

The fastest way to go from idea to shipped code. One agent thinks, then many agents build.

1. Research — Single agent, deep thinking

Use a persona to explore the problem space and draft your planning prompt:

```

/architect redesign the auth system

```

This produces artifacts (PR-FAQs, PRDs, ADRs) that feed the next step.

2. Plan — Parallel exploration

```

/swarm-plan implement user notifications per the PRD

```

Launches 3-6 explorer agents in parallel. They research existing patterns, dependencies, and constraints simultaneously, then produce a decomposed plan tracked as Beads.

3. Execute — Parallel implementation

```

/swarm-execute

```

Picks up the planned work and fans out across multiple builder agents. Each worker claims a task, implements it, runs quality gates (tests, lint, types, build), and pushes to remote. Up to 8 workers in parallel.

4. Review — Adversarial, multi-pass

```

/swarm-review # run 2-3 times

```

Each pass launches 5 parallel reviewers covering different perspectives: security (OWASP Top 10), performance, architecture, test coverage, and code quality. Repeat until clean. Then ship the PR.

Full cycle

```

/architect → /swarm-plan → /swarm-execute → /swarm-review (2-3x) → PR

```

Eight specialized agent types, each tuned for cost and capability:

| Worker | Model | Use |

|--------|-------|-----|

| worker-explorer | Haiku | Fast codebase search |

| worker-builder | Sonnet | Implementation |

| worker-reviewer | Sonnet | Code review |

| worker-tester | Sonnet | Test writing |

| worker-researcher | Sonnet | Web research |

| worker-architect | Opus | Complex design decisions |

| worker-security | Sonnet | Security analysis |

| worker-refactor | Sonnet | Code cleanup |

Haiku for read-only exploration. Sonnet for most work. Opus only for hard architectural calls.

For single-agent work, ideation, and research:

```

/architect /builder /qa-engineer

/security-auditor /ui-ux-designer /code-check

/swarm-plan /swarm-execute /swarm-review

```

Personas produce artifacts in artifacts/ that chain naturally into swarm workflows.

• 65+ Skills — Auto-suggested based on context (designing-systems, debugging, react-patterns, ...)
• Beads — Lightweight issue tracking that coordinates swarm workers via git
• MCP Servers — Browser debugging, GitHub integration, structured reasoning, library docs
• /code-check — Holistic codebase audit for SOLID, DRY, consistency

1. Open CLAUDE.md and add your build/test commands
2. Required: Edit .claude/rules/tech-strategy.md to match your tech stack
3. Try the swarm workflow on your next feature

• [Installation details](docs/getting-started.md)
• [Multi-agent swarms](docs/swarm.md)
• [All personas](docs/personas.md)
• [Skills reference](docs/skills.md)
• [Beads issue tracking](docs/beads.md)
• [MCP servers](docs/mcp-servers.md)
• [Customization](docs/customization.md)
• [Hooks](docs/hooks.md)
• [Handoffs](docs/handoffs.md)