By adihebbalae
Multi-agent AI coding orchestrator. Manager coordinates, Engineer builds, Security audits. Install once, works across all your Claude Code projects.
2-minute live demo of Attacca. Shows the full PRD → agents → build pipeline using a sample project. Writes no files. Use this when showing Attacca to someone new.
Hand off an architectural decision to the Consultant agent by TASK-ID. Use when Engineer is blocked after 3 attempts, or for complex tradeoff decisions.
Hand off a UI/UX task to the Designer agent by TASK-ID. Reads task title and context automatically.
Hand off a task to the Engineer agent by TASK-ID. Reads task title and context automatically.
Hand off a research task to the Researcher agent by TASK-ID. Reads task title and context automatically.
Deep architectural reasoning agent (expensive — use sparingly). Invoke when Engineer is blocked after 3 attempts, for architecture decisions affecting multiple domains, cross-cutting changes, CRITICAL security findings, or conflicting requirements with non-obvious tradeoffs.
UI/UX reviewer and design spec generator. Use for reviewing visual designs, writing component specs, accessibility audits, and design-to-code guidance. Does not write code.
Code implementation agent. Use for building features, fixing bugs, running tests, and committing code. Works autonomously — does not ask the user questions when running as a subagent.
Project orchestrator and user's primary contact. Coordinates all agents, plans tasks, manages handoffs. Use for: starting features, planning work, reviewing progress, coordinating handoffs, making architectural decisions.
SEV 1 emergency incident responder. Invoke via /hotfix for: app won't start, 500 errors on critical flows, broken deploy pipelines, complete test suite failures, database connection failures. Has autonomous deployment authority. DO NOT use for SEV 2+ (degraded but not down) — use engineer instead.
On-demand code review checklist. Use when: reviewing code changes, PR review, self-review before committing, checking code quality after implementation. Covers readability, correctness, performance, security, and maintainability.
Pre-push quality gate checklist. Use when: before any git push, before opening a PR, after implementing a full feature, when Manager says 'run quality gate'. Runs lint, type-check, tests, and a security scan as a single sequential gate. Fails fast — does not proceed past a failing stage.
On-demand security audit checklist based on OWASP Top 10. Use when: pre-push security review, auditing new features for vulnerabilities, checking authentication/authorization flows, reviewing API security, validating input sanitization, dependency vulnerability scanning.
TDD workflow skill enforcing RED → GREEN → REFACTOR. Use when: implementing any feature test-first, writing unit or integration tests, building with a test-heavy stack (Jest, Vitest, pytest, RSpec, etc.). Prevents writing implementation before a failing test exists.
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
attacca (Italian) — proceed to the next movement without pause.
PRD → Project. Paste a product requirements doc, answer a few questions, and a multi-agent system plans, builds, tests, and commits the code — while you review.
Orchestration layer for any AI coding tool: GitHub Copilot, Cursor, Cline, Windsurf, Claude Code, Codex CLI, Gemini CLI, and Google Antigravity. Manager coordinates, Engineer implements, Security audits before every push. All modes share the same state files.
npx create-attacca my-project
cd my-project
The CLI asks which tools you use, your LLM backend (cloud/local/hybrid), agent complexity, and skill packs — then generates only the files you need.
gh repo create my-project-name --template adihebbalae/Attacca --public --clone
cd my-project-name
Then:
/init-project with your PRDPaste your PRD. Research runs automatically. That's it — PRD → Research → Project.
Note:
/init-projecthandles the.gitignore.projectrename automatically.
| Tool | Config Format | Files |
|---|---|---|
| GitHub Copilot | .github/agents/*.agent.md + .github/copilot-instructions.md | 7+ agent files, prompts, skills |
| Cursor | .cursor/rules/*.mdc (YAML frontmatter) | 7 rule files with alwaysApply / agent-decision activation |
| Cline | .clinerules/*.md (optional paths: frontmatter) | 6 rule files with path-conditional activation |
| Windsurf | .windsurfrules (single file) | 1 concatenated protocol file |
| Claude Code CLI | CLAUDE.md + .claude/agents/*.md | Bootstrap + subagent definitions + settings |
| Codex CLI | AGENTS.md | Single bootstrap file (manual handoff mode) |
| Gemini CLI | GEMINI.md + .gemini/settings.json | Bootstrap + hooks config |
| Google Antigravity | .agents/rules/*.md + .agents/workflows/*.md | 4 rules + 3 workflows (native skill format match) |
Cross-compatibility note: AGENTS.md is read natively by Cursor, Cline, and Zed. Zed also reads .cursorrules, .windsurfrules, .clinerules, copilot-instructions.md, CLAUDE.md, and GEMINI.md.
When you run /init-project, Manager asks two setup questions to adapt the workflow:
| Question | Options | Impact |
|---|---|---|
| Do you have a CLI agent? | Yes / No | No CLI? Everything routes through GitHub Copilot (160k context). Have CLI (Claude Code, Codex, or Gemini)? Unlock Complex Project Mode for 20+ file projects. |
| What's your budget? | Free tier / Paid / TBD | Free? Manager adds a research task to find free deployment options. Paid? Use production-grade tools from day one. |
Why? This ensures the boilerplate works for everyone: Copilot-only users, budget-conscious teams, and power users with CLI access. Your answers are saved in .agents/state.json and used by Manager for routing decisions.
Want to change your answers later? Run /setup-budget anytime.
Current version: v3.0.0 — See CHANGELOG.md for full version history and upgrade notes.
To update an existing project to the latest boilerplate version, run /update-boilerplate from the Manager agent.
This boilerplate works in eight modes. All share the same state files.
Agents live in .github/agents/*.agent.md. Open the Copilot chat panel, select Manager, and start.
github.copilot.chat.claudeAgent.enabled: true in VS Code settings..agents/handoff.md. You copy them to the target agent using /handoff-to-[agent] prompts.Agents live in .claude/agents/*.md. Install Claude Code, run claude from the project root. Claude reads CLAUDE.md as a bootstrap and uses .claude/agents/ for subagent definitions.
npm install -g @anthropic-ai/claude-code
claude
npx claudepluginhub adihebbalae/attacca --plugin attacca11 agents, 35 skills, 18 commands, 9 hooks — spec-driven multi-agent orchestration for Claude Code, with optional cross-device semantic memory.
Complete project development toolkit: 23 agents, 23 slash commands, 29 lifecycle hooks, and 69 reusable skills for Claude Code workflows
Autonomous development methodology: PRD interviews → agent execution → automated review
Project discovery, planning, and quality analysis system with parallel agents
Persona-driven AI development team: orchestrator, team agents, review agents, skills, slash commands, and advisory hooks for Claude Code
GSD Core is a meta-prompting, context engineering, and spec-driven development system for AI coding agents.