cookiy

# Cookiy Skill Turn Claude Code, Codex, Cursor, VS Code / GitHub Copilot, Windsurf, Cline, OpenClaw, Manus, and other MCP clients into AI user-research operators that can create studies, run interviews, recruit participants, and deliver report links from plain-English prompts. This repository now has two layers: - a protected production surface for the current public `cookiy` skill and MCP install flow - an additive open library of prompts, references, templates, and examples for user research The install and indexing surface stays stable while the library grows around it. The public installer package now acts as a bootstrap CLI: 1. on supported clients, it installs a local Cookiy skill copy 2. then it configures MCP 3. on unsupported clients, it falls back to MCP-only setup ## What You Get In This Repo Today This repo already includes concrete user-research materials, not just install docs. | Material type | What is included now | What it is for | |---|---|---| | Installable runtime skill | `cookiy` skill + MCP install surface | Connect an agent to Cookiy and run the live study/interview/recruit/report workflow | | Public installer package | `packages/cookiy-mcp/` | Source for the `cookiy-mcp` bootstrap CLI and Homebrew build artifacts | | Study-brief prompts | `prompts/study-briefs/discovery-study.md` | Turn a vague founder or product question into a stronger study brief | | Interview-guide prompts | `prompts/interview-guides/problem-exploration.md` | Draft better exploratory interview sections, probes, and evidence goals | | Synthesis prompts | `prompts/synthesis/evidence-first-synthesis.md` | Convert raw notes, transcripts, survey comments, or support inputs into findings | | Stakeholder readout prompts | `prompts/stakeholder-readouts/executive-summary.md` | Turn findings into concise founder/PM/stakeholder summaries | | Research method references | `references/methods/research-method-selector.md`, `references/methods/thematic-analysis.md` | Choose the right method and synthesize qualitative evidence better | | Reusable templates | `references/templates/research-plan-template.md`, `references/templates/insight-card-template.md` | Standardize research planning and finding documentation | | Book-ingestion guidance | `references/books/README.md` | Turn books into agent-usable notes without uploading copyrighted content | | Worked examples | `examples/study-briefs/saas-onboarding-friction.md`, `examples/readouts/mobile-checkout-friction.md` | Show what good study briefs and research readouts look like | | Repo governance docs | `docs/PLATFORM_COMPATIBILITY.md`, `docs/CONTENT_POLICY.md`, `docs/THIRD_PARTY_ATTRIBUTIONS.md`, `docs/ROADMAP.md` | Keep public platform compatibility stable and define how the library should grow | ## Materials By Use Case If you are here for a specific kind of material, start here: - To plan a study: - `prompts/study-briefs/discovery-study.md` - `references/templates/research-plan-template.md` - `examples/study-briefs/saas-onboarding-friction.md` - To maintain the public installer package: - `packages/cookiy-mcp/package.json` - `packages/cookiy-mcp/README.md` - `packages/cookiy-mcp/scripts/generate-homebrew-formula.mjs` - `docs/contract-source.json` - To write or improve interview guides: - `prompts/interview-guides/problem-exploration.md` - `references/methods/research-method-selector.md` - To analyze research data: - `prompts/synthesis/evidence-first-synthesis.md` - `references/methods/thematic-analysis.md` - `references/templates/insight-card-template.md` - To write readouts for founders or PMs: - `prompts/stakeholder-readouts/executive-summary.md` - `examples/readouts/mobile-checkout-friction.md` - To build a learning library safely: - `references/books/README.md` - `docs/CONTENT_POLICY.md` - `docs/THIRD_PARTY_ATTRIBUTIONS.md` ## 30-Second Demo ```bash # Install Cookiy for your AI client npx cookiy-mcp --client codex -y # Then ask your agent: "Create a 6-participant checkout-abandonment study in English." "Run 3 simulated interviews with price-sensitive mobile shoppers." "Share the report link as soon as a preview is ready." ``` What happens: - Cookiy creates the study and discussion guide inside your agent workflow. - It runs or monitors interviews and recruitment through the MCP tool chain. - It returns transcripts, status updates, and share links without leaving chat. ## 3 Real Examples | Use case | Ask your agent | What Cookiy does | |---|---|---| | Ecommerce checkout drop-off | "Create a 6-participant study to understand why mobile shoppers abandon checkout after shipping costs appear." | Creates the study, generates the guide, and highlights sample size, interview mode, and duration for review. | | SaaS onboarding friction | "Run 5 simulated interviews with first-time admins setting up SSO for the first time." | Queues AI interviews, polls status, and returns transcripts for early signal before spending on real participants. | | Concept validation with real users | "Recruit bilingual parents in the US for 8 short interviews about a children's learning app, then share the report link when ready." | Previews targeting and pricing, launches recruitment after confirmation, tracks progress, and returns the report share link when available. | ## Quick Install ### Claude Code Install as a plugin directly in Claude Code: ```bash claude plugin add cookiy-ai/cookiy-skill ``` Or install the MCP server standalone: ```bash npx cookiy-mcp --client claudeCode -y ``` ### Codex ```bash npx skills add cookiy-ai/cookiy-skill ``` Or install the MCP server directly: ```bash npx cookiy-mcp --client codex -y ``` ### OpenClaw ```bash clawhub install cookiy # Skill only, not MCP ``` This installs the Cookiy skill only. To connect OpenClaw to Cookiy's live MCP tools, install the MCP server separately: Or install the MCP server directly: ```bash npx cookiy-mcp --client openclaw -y ``` ### VS Code / GitHub Copilot ```bash npx cookiy-mcp --client vscode -y ``` ### Windsurf ```bash npx cookiy-mcp --client windsurf -y ``` ### Cline ```bash npx cookiy-mcp --client cline -y ``` ### Manus ```bash npx cookiy-mcp --client manus -y ``` This writes a resumable headless OAuth helper bundle under `~/.mcp/<server>/` for Manus-style sandbox environments. ### Cursor Install from the [Cursor Marketplace](https://cursor.com/marketplace), or manually: ```bash npx cookiy-mcp --client cursor -y ``` ### Any platform with Node.js ```bash npx cookiy-mcp -y ``` The installer auto-detects installed AI clients. On supported clients, it installs a local Cookiy skill copy first and then configures MCP. Other clients fall back to MCP-only setup. ### macOS with Homebrew ```bash brew install cookiy-ai/tap/cookiy && cookiy -y ``` ## Security This repository contains both skill content and an installer that writes local MCP configuration for supported clients. That means automated platform scanners may flag expected behaviors such as OAuth flows, client config updates, helper scripts, and local skill installation. For a concrete description of the intended network access, local file writes, and credential model, see [SECURITY.md](SECURITY.md). ## Verification Matrix Validation note: rows below reflect direct `cookiy-mcp` CLI dry-runs against the live installer contract. | Client | Direct install command | Local validation status | Last verified | |---|---|---|---| | Claude Code | `npx cookiy-mcp --client claudeCode -y` | CLI dry-run verified, client detected locally | 2026-03-17 | | Codex | `npx cookiy-mcp --client codex -y` | CLI dry-run verified, client detected locally | 2026-03-17 | | Cursor | `npx cookiy-mcp --client cursor -y` | CLI dry-run verified, client detected locally | 2026-03-17 | | GitHub Copilot / VS Code | `npx cookiy-mcp --client vscode -y` | CLI dry-run verified, client detected locally | 2026-03-17 | | OpenClaw | `npx cookiy-mcp --client openclaw -y` | CLI dry-ru