Project enabler — takes a raw idea and transforms it into a fully scaffolded project with research-backed design artifacts (ERD, schema.sql, PRD, API contract, architecture) and a personalized .claude/ ecosystem. Integrates deep market research via Perplexity API and monetization strategy via /monetize. Use this skill whenever the user pitches a new project idea, wants to start a project from scratch, says "kickstart", "new project idea", "I have an idea", "let's build", "project enabler", or any variation of going from idea to implementation-ready. Always use this skill before /bootstrap when starting from zero.
From cksnpx claudepluginhub cardinalconseils/claude-starter --plugin cksThis skill is limited to using the following tools:
README.mdreferences/ai-glossary.mdreferences/phase-banners.mdreferences/validation-and-state.mdworkflows/auto-chain.mdworkflows/brand.mdworkflows/compose.mdworkflows/design.mdworkflows/handoff.mdworkflows/intake.mdworkflows/research.mdworkflows/stack-selection.mdEnables AI agents to execute x402 payments with per-task budgets, spending controls, and non-custodial wallets via MCP tools. Use when agents pay for APIs, services, or other agents.
Compares coding agents like Claude Code and Aider on custom YAML-defined codebase tasks using git worktrees, measuring pass rate, cost, time, and consistency.
Designs and optimizes AI agent action spaces, tool definitions, observation formats, error recovery, and context for higher task completion rates.
This skill is loaded into kickstart agents via the skills: kickstart frontmatter field.
It provides domain expertise — not execution instructions. Agents read workflow files
in workflows/ for step-by-step process via progressive disclosure.
Takes a raw project idea through guided discovery, optional market research & monetization
analysis, then generates design artifacts and hands off to /bootstrap to wire up the
full .claude/ ecosystem.
/kickstart → ideate? → intake → compose → stack → research? → monetize? → brand? → design → handoff → /cks:new → discover
Each phase is independently resumable. The command reads .kickstart/state.md to determine
where to resume and dispatches the appropriate agent.
| Phase | Agent | Required? | Output |
|---|---|---|---|
| 0 — Ideate | kickstart-ideator | Optional | .kickstart/ideation.md |
| 1 — Intake | kickstart-intake | Yes | .kickstart/context.md |
| 1b — Compose | kickstart-intake | Yes | .kickstart/manifest.md |
| 1c — Stack | kickstart-intake | Yes | .kickstart/stack.md |
| 2 — Research | deep-researcher | Optional | .kickstart/research.md |
| 3 — Monetize | monetize-discoverer | Optional | .monetize/ |
| 4 — Brand | kickstart-brand | Optional | .kickstart/brand.md |
| 5 — Design | kickstart-designer | Yes | .kickstart/artifacts/ |
| 6 — Handoff | kickstart-handoff | Yes | CLAUDE.md, .prd/, scaffold |
Phase 1b (Compose) is ALWAYS run after Intake. Phases 2 (Research), 3 (Monetize), and 4 (Brand) are OPTIONAL for the USER but the QUESTION is MANDATORY for YOU.
You MUST call AskUserQuestion at each gate. You MUST NOT:
The user's explicit response is what drives skip/proceed — never your inference.
Failure mode to avoid: Claude sees no API key or sees .prd/ already exists and decides "they probably don't want this" and skips the AskUserQuestion call entirely. This is WRONG. Always ask. Every time. No exceptions.
STOP RULE: After completing any phase (including marking it skipped), you MUST:
.kickstart/state.md BEFORE displaying the completion banner.kickstart/state.md BEFORE starting the next phaseIf you do not update the state file, the resume system breaks and phases get re-run or skipped incorrectly on the next invocation.
For the full state file template and validation rules per phase, read references/validation-and-state.md.
Display a progress banner at the start of every phase showing all phases and their status:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
KICKSTART ► {PHASE_NAME}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[0] Ideate {✅ done | ▶ current | ○ pending | ⊘ skipped}
[1] Intake {✅ done | ▶ current | ○ pending | ⊘ skipped}
[1b] Compose {✅ done | ▶ current | ○ pending | ⊘ skipped}
[1c] Stack {✅ done | ▶ current | ○ pending | ⊘ skipped}
[2] Research {✅ done | ▶ current | ○ pending | ⊘ skipped}
[3] Monetize {✅ done | ▶ current | ○ pending | ⊘ skipped}
[4] Brand {✅ done | ▶ current | ○ pending | ⊘ skipped}
[5] Design {✅ done | ▶ current | ○ pending | ⊘ skipped}
[6] Handoff {✅ done | ▶ current | ○ pending | ⊘ skipped}
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
After each phase completes, display:
[N] {Phase Name} ✅ done
Output: {artifact file path(s)}
Read .kickstart/manifest.md to determine mode:
.kickstart/artifacts/.kickstart/artifacts/shared/.kickstart/artifacts/sp-{NN}-{name}/| Phase | Requires |
|---|---|
| Ideate | Nothing (entry point) |
| Intake | Nothing (or .kickstart/ideation.md if ideation was run) |
| Compose | .kickstart/context.md |
| Stack | .kickstart/context.md + .kickstart/manifest.md |
| Research | .kickstart/context.md + PERPLEXITY_API_KEY (or deep-research sources) |
| Monetize | .kickstart/context.md (Perplexity optional — falls back to WebSearch) |
| Brand | .kickstart/context.md (uses Canva MCP, WebFetch, or manual Q&A) |
| Design | .kickstart/context.md + .kickstart/manifest.md (research/monetize/brand optional but consumed if present) |
| Handoff | .kickstart/manifest.md + design artifacts per sub-project |
| File | Purpose |
|---|---|
.kickstart/ideation.md | Refined idea pitch from brainstorming (if ideation was run) |
.kickstart/state.md | Progress tracker — resume point on interruption |
.kickstart/context.md | Full idea context from intake Q&A |
.kickstart/manifest.md | Project composition — sub-projects, dependencies, build order |
.kickstart/stack.md | Technology stack decisions — frameworks, database, hosting, CI/CD |
.kickstart/research.md | Market research with citations (if opted in) |
.kickstart/brand.md | Brand guidelines — colors, typography, voice, UI prefs (if opted in) |
.kickstart/artifacts/ | Design artifacts (ERD, schema, PRD, API, Architecture, Feature Roadmap) |
.monetize/* | Monetization artifacts (if opted in) |
.prd/PROJECT-MANIFEST.md | Manifest copy for feature lifecycle (created during handoff) |
| Variable | Required By | Phase | Fallback |
|---|---|---|---|
PERPLEXITY_API_KEY | Research (standard), Monetize | 2, 3 | WebSearch (Claude built-in) |
If the key is missing, research and monetize phases automatically fall back to WebSearch.
| Failure | Behavior |
|---|---|
PERPLEXITY_API_KEY missing | Fall back to WebSearch, note source in output |
| Perplexity rate limit | Retry once, then save partial results and continue |
| User abandons mid-intake | Save partial context + state.md, next run offers to resume |
/bootstrap not available | Generate artifacts only, print manual bootstrap instructions |
| Phase validation fails | Retry once, then ask user what to do |
| Scaffold build fails | Report error, save what was created, continue to next sub-step |
During Q&A, when the user mentions AI/tech concepts, briefly explain using the glossary
in references/ai-glossary.md:
> **{Term}** — {one-line definition from glossary}
Only surface when natural — don't force explanations.
| File | When to Read |
|---|---|
workflows/intake.md | Phase 1 — intake Q&A steps |
workflows/compose.md | Phase 1b — sub-project identification |
workflows/stack-selection.md | Phase 1c — technology stack & platform selection |
workflows/brand.md | Phase 4 — brand extraction steps |
workflows/design.md | Phase 5 — artifact generation steps |
workflows/handoff.md | Phase 6 — scaffolding steps |
workflows/auto-chain.md | After Phase 6 — feature lifecycle handoff |
references/validation-and-state.md | Validation rules + state file format |
references/phase-banners.md | Sub-step validation banners for Phase 5/6 |
references/ai-glossary.md | During intake Q&A — surface relevant definitions |
This skill ships with opinionated defaults. Review and adapt to your needs:
references/validation-and-state.mdworkflows/design.mdworkflows/auto-chain.mdreferences/ai-glossary.md| Rationalization | Reality |
|---|---|
| "Let me just start coding" | 30 minutes of planning saves 3 hours of rework. Kickstart produces the design artifacts coding needs. |
| "I'll figure out the architecture as I go" | Architecture decisions made under implementation pressure are consistently worse than planned ones. |
| "We don't need a design system yet" | Inconsistent UI from day one creates technical debt that compounds with every new screen. |
| "Research is a waste of time" | Building on wrong assumptions is the real waste. Research validates before you invest implementation time. |