Planning & Workflow

Pressure-test a design by walking its decision tree one question at a time

Run quality gates, review staged changes and write a conventional commit

Drive a change through red, green and refactor with a failing test first

Inspect context and write a markdown plan to a folder without running the work

Run a lightweight dev workflow with clarified requirements and 90% test coverage

Build features through five phases: understand, clarify, design, implement, complete

Work through a specify, plan, act, review and vault workflow with a spec gate

Stage changes, split them into logical commits and write clear commit messages

Write sectioned implementation plans from research, interviews and multi-model review

Extract engineering lessons from recent git changes and write them down

Clean up stale git worktrees and their old local branches on request

Close the GitHub issue for a finished branch and remove its worktree and local branch

Write clean Git commit messages and decide how to split changes before a PR

Manage GitHub issues, pull requests and review threads through the GitHub MCP

Create a GitHub issue branch and matching git worktree, then open it in Cursor

Decide whether an AI coding task needs a spec, several agents or a checklist

Take an app idea through validation to a clean first version that ships

Convert a PRD into a beads epic with child tasks for ralph-tui

Convert a PRD into beads-rust tasks for ralph-tui using the br CLI

Convert a PRD into prd.json task files for ralph-tui execution

Summarize open pull requests into a blockers first standup digest

Isolate feature work in its own branch or worktree and merge it cleanly

Make build-versus-buy, tech stack and engineering capacity calls

Turn a spec into a step-by-step plan that another session can execute

Design system boundaries and record architecture decisions and their trade-offs

Explore the problem and approach options before any code is written

Solve multi-step problems by defining done, gathering evidence and verifying outcomes

Compare npm packages side by side on downloads, bundle size, stars and CVEs

Find maintained open source alternatives to paid SaaS tools and commercial APIs

Recommend which architecture commands to run next for a new or in-flight project

Build a project's architecture artefacts in parallel from a YAML recipe with resume

Execute a written implementation plan in a separate session with review checkpoints

Combine unrelated ideas to surface new properties and approaches

Flip core assumptions to reveal hidden constraints and alternative approaches

Spot patterns across three or more domains to find universal principles

Break through stuck points with inversion, simplification and pattern recognition

Test ideas at extreme scales to expose truths hidden at normal sizes

Find one insight that removes several components from the design

Pick the right problem-solving technique based on how you are stuck

Manage Forgejo issues, pull requests, releases and runners with a gh-style CLI

Find duplicate issues, chase pending reviews and search GitHub data offline

Download GitHub folders without cloning, verify copies and re-sync only changes

Read GitLab merge requests, diffs, issues and pipelines from the terminal

Write a markdown issue spec with a concise implementation plan for a task

Explore intent, requirements and design before building a feature or change

Run real verification commands and confirm output before claiming work passes

Turn a multi-step spec or requirements into a concrete implementation plan

Automate GitHub issues, pull requests, branches and CI through Rube MCP

Search for existing tools, libraries and patterns before writing custom code