shapeup

Shape Up Skills

An agent skill that guides you through the Shape Up product planning methodology. Instead of generating documents, it conducts structured conversations that pull information out of your head, challenge your assumptions, and produce clear artifacts.

Works with any compatible agent and for solo builders and teams alike.

See CHANGELOG.md for versioned release notes.

What it does

The skill walks you through the core Shape Up phases plus optional upstream phases:

• Explore (optional) — Diverge. Map the problem space, find the people, understand the landscape, and surface multiple candidate wedges before narrowing.
• Evidence (optional) — Test. Run lightweight interviews, smoke tests, concierge tests, fake doors, or other cheap experiments when a promising wedge still needs real-world signal before it deserves appetite.
• Frame — Define the real problem, who it affects, and how much time it's worth (appetite).
• Shape — Find the elements of a solution at fat-marker fidelity with explicit boundaries.
• De-risk — Stress-test the concept for rabbit holes, unknowns, and scope bombs.
• Pitch — Synthesize everything into a decision document for the build phase.

Each phase produces a markdown artifact in your project's shaping/ directory. The skill tracks progress via YAML frontmatter, so you can resume across sessions.

Current Feature Set

• Six-phase planning flow: optional Explore and Evidence upstream phases, followed by Frame, Shape, De-risk, and Pitch
• Explicit phase handoffs: documented checks for when to stay broad, when to gather evidence, and when a problem is ready to frame
• Resumable artifacts: every artifact carries frontmatter plus Open Questions and Next Best Question sections so the skill can pick up cleanly in later sessions
• Evidence workflow: lightweight support for interviews, smoke tests, concierge tests, fake doors, and other cheap validation work before appetite is committed
• Completed example runs: four sample projects showing tone, pacing, phase handoffs, and finished artifacts
• Installable plugin packaging: plugin manifest, marketplace metadata, and versioned release notes in CHANGELOG.md

Install

Claude Code plugin (recommended)

# Add the marketplace and install
/plugin marketplace add onliminal/shapeup-skills
/plugin install shapeup@shapeup-skills

Or test locally:

claude --plugin-dir ./path/to/shapeup-skills

Claude Code standalone skill

Copy or symlink the skill directory:

# Personal (available across all projects)
cp -r skills/shapeup ~/.claude/skills/shapeup

# Project-specific
cp -r skills/shapeup .claude/skills/shapeup

See the Claude Code skills documentation for more details.

Codex CLI

cp -r skills/shapeup ~/.codex/skills/shapeup

See the Agent Skills specification for the standard skill format.

OpenCode

Clone the repo into the OpenCode skills directory:

git clone https://github.com/onliminal/shapeup-skills.git ~/.opencode/skills/shapeup-skills

OpenCode auto-discovers all SKILL.md files under ~/.opencode/skills/. No config changes needed — restart OpenCode to activate.

Other agents

Any agent that supports the Agent Skills format (Cursor, Gemini CLI, VS Code Copilot, Goose, Roo Code, etc.) can use this skill — consult your agent's documentation for the skills directory path.

Usage

The skill triggers automatically when you mention shaping, framing, pitching, appetite, or related concepts. You can also start at a specific phase:

I have a product idea I want to explore
Help me design an evidence plan for this wedge
Frame this problem for me
Help me shape a solution
De-risk this concept
Write a pitch

The skill detects prior work in shaping/ and picks up where you left off.

When a project resumes, the skill reads the existing artifacts, surfaces unresolved questions, and by default continues with the recorded next-best question.

Examples

The skill now includes completed sample runs under skills/shapeup/examples/:

• new-product — a full upstream-to-pitch example for a new product idea
• internal-tool — a tighter feature/planning flow for an internal ops tool
• customer-request — a solution-first customer request reframed into a real problem
• ai-workflow — an AI/workflow concept with Explore and Evidence before Framing

Each example includes:

• README.md — what the example is meant to teach
• transcript.md — an abridged sample conversation showing tone, pacing, checkpoints, and phase handoffs
• shaping/{project-slug}/... — the completed artifacts from that run