Founder Mode

Founder Mode — End-to-End Product Builder

From idea to launched product. Not just code — a product.

How It Works

/founder-mode "Build a CLI that converts markdown to slides"
        │
        ▼
  Phase 0: Bootstrap ─── config.json + state.json + mode detection
        │
        ▼
  Phase 1: Idea Evaluation ─── /architect dispatches 6 evaluators
        │
        ▼
  ══ GATE 1 ══ Go/No-Go decision ─── [continue|reject|iterate|edit]
        │
        ▼
  Phase 2: Architecture ─── system design, components, tech stack
        │
        ▼
  Phase 3: TDD Planning ─── decompose into test-first chunks
        │
        ▼
  Phase 4: Implementation ─── chunk-with-summary TDD loop
        │
        ▼
  ══ GATE 2 ══ Implementation review ─── [continue|iterate|stop|edit]
        │
        ▼
  Phase 5: E2E Testing ─── strategy by project type
        │
        ▼
  Phase 6: Verification ─── false positives + spec-to-test gap + arch drift
        │
        ▼
  ══ GATE 3 ══ Verification review ─── [continue|iterate|stop|edit]
        │
        ▼
  Phase 7: Post-Run Analysis ─── lessons + recommendations
        │
        ▼
  Phase 8: GTM Draft ─── README + pitch + tweets + PH + CHANGELOG
        │
        ▼
  DONE ─── all artifacts in .founder/

Phases

| Phase | What It Does | Agent | |-------|-------------|-------| | 0 - Bootstrap | Config generation, mode detection, codebase scan (existing projects), lesson loading | bootstrapper.md | | 1 - Evaluation | Dispatches 6 parallel evaluators via /architect, extracts acceptance criteria | evaluator.md | | 2 - Architecture | Generates system design (new) or reverse-engineers from codebase (existing) | architect-gen.md | | 3 - TDD Planning | Decomposes architecture into ordered test-first chunks | planner.md | | 4 - Implementation | TDD per chunk with cross-chunk summaries, supports re-entry for fixes | implementer.md | | 5 - E2E Testing | Generates and runs E2E tests matched to acceptance criteria | e2e-runner.md | | 6 - Verification | False-positive detection, spec-to-test gap analysis, architecture drift check | verifier.md | | 7 - Analysis | Post-run pattern analysis, lesson extraction, recommendations | analyzer.md | | 8 - GTM Draft | README, pitch, tweet thread, Product Hunt listing, CHANGELOG | gtm-drafter.md |

Human Gates

Three checkpoints where you make Go/No-Go decisions:

• Gate 1 (Post-Eval): Review the idea score. Continue, reject, or iterate with refined framing.
• Gate 2 (Post-Implementation): Review test results and known issues. Continue or iterate to fix.
• Gate 3 (Post-Verification): Review verification verdict. Continue to GTM or iterate to fix issues.

Each gate supports: [continue], [iterate], [stop], [edit]

Iteration starts with a cap of 3 (configurable). When reached, you're prompted to extend or move on.

Config

Generated at .founder/config.json during bootstrap. Key settings:

• mode: "new" or "existing" (auto-detected)
• maxIterations: iteration cap (default 3, extendable at gates)
• gates: per-gate behavior — "pause", "auto", or "skip"
• phases: enable/disable individual phases
• tdd: strictness and coverage threshold
• gtm: enable/disable individual GTM artifacts

Edit the config before saying go to customize the pipeline.

Modes

• New Idea: Full pipeline from Phase 0 to Phase 8.
• Existing Project: Pass a path (/founder-mode ./my-project). Skips eval, reverse-engineers architecture, plans around existing tests, extends code, respects existing test framework.

Dependencies

• Required: architect skill (for Phase 1 evaluation)
• Optional: learn-by-mistake (error memory), post-run-analysis (enhanced Phase 7)

Usage

/founder-mode "Build a task management API with natural language input"
/founder-mode "./my-existing-project"
/founder-mode resume

Crash Recovery

State is persisted to .founder/state.json after every phase and gate decision. If your session crashes, run /founder-mode resume to pick up exactly where you left off — including mid-gate crashes.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: JuanMarchetto
• Source: JuanMarchetto/agent-skills
• License: MIT
• Homepage: https://github.com/JuanMarchetto/agent-skills

Install and usage instructions live in the source repository linked above.