autoresearch
Autonomously optimize any Claude Code skill by running it repeatedly, scoring outputs against binary evals, mutating the prompt, and keeping improvements. Based on Karpathy's autoresearch methodology. Use when: optimize this skill, improve this skill, run autoresearch on, make this skill better, self-improve skill, benchmark skill, eval my skill, run evals on. Outputs: an improved SKILL.md, a results log, and a changelog of every mutation tried.
check-env-keys
Verify which environment variable keys are present in .env files or shell environment without exposing their values. Use when you need to check env configuration or compare keys between .env files.
agent-ready-repo-setup
|
react-hook-form
Build forms in React using React Hook Form and Zod validation. Use when creating forms with useForm, Controller, or useFieldArray.
task-estimation
Provides task estimation before starting work. Use when planning new features, refactoring, bug fixes, or any non-trivial changes. Triggered by keywords like estimate, plan, complexity, tokens, approach.
tanstack-form
Build forms in React using TanStack Form and Zod validation. Use when creating forms, handling form validation, or working with form fields.
component-organization
Enforces single-component-per-file pattern for React feature pages. Use when creating new pages, refactoring page.tsx files, or when components are mixed in a single file.
feature-scaffolding
Scaffolds new feature pages following project patterns. Use when creating new pages, features, routes, or when user asks to set up a new section. Works with component-organization skill for structure.
supabase-setup
Use when setting up Supabase, creating database tables, writing RLS policies, configuring Auth (especially Apple Sign-In), creating storage buckets, writing Edge Functions in TypeScript/Deno, or running migrations. Triggers on "Supabase setup", "RLS policy", "Edge Function", "database schema", "storage bucket", "Apple Sign-In auth".
writing-plans
Use when you have a spec or requirements for a multi-step task, before touching code
verification-before-completion
Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
testing-flexible
Use when implementing features to write effective tests. Encourages test-first but allows prototyping.
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
subagent-driven-development
Use when executing implementation plans with independent tasks. Features 3-stage review: spec compliance, code quality, then UX.
swift-ios-app
Use when creating iOS apps, setting up Xcode projects, designing app architecture, implementing SwiftUI views, using SwiftData models, adding Swift 6 concurrency with actors, or integrating Supabase backend. Triggers on "new iOS app", "Swift architecture", "SwiftData setup", "actor pattern", "iOS project structure".
receiving-code-review
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
dispatching-parallel-agents
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
brainstorming
MUST use before building features, creating components, adding functionality, or designing new behavior. Use when user says 'build', 'create', 'add feature', 'implement', 'design', or describes something new to build. Explores requirements through questions before writing code.
open-brush-plugin-skill
Create and modify Lua plugins for Open Brush with full API documentation access. Use when the user wants to create Open Brush plugins, work with Lua scripts for Open Brush, or asks about Open Brush scripting API, Lua functions, or plugin development.
knowledge-base
专业的知识库管理系统，旨在解决“知识诅咒”和认知偏差问题。通过显式化隐性知识、扫描代码提取领域概念、整合行业最佳实践，构建结构化的 Markdown 知识库。
skill-learner
Reviews conversation context to identify learnings and patterns that should be captured in skills. Use when the user asks to review the session for skill creation or updates.
software-testing
Software testing guidance with decision tree. Use when: (1) Writing or improving tests, (2) User asks about testing approaches, (3) Need guidance on test scope or assertions, (4) Working with specific test frameworks. Routes to specialized testing strategies via reference files.
emacs-elisp-debugging
Strategies for debugging Emacs Lisp code, especially missing parentheses and Common Lisp compatibility issues. Use when encountering mysterious elisp errors or when working with literate elisp in org-mode files.
writing-elisp
Comprehensive guidance for writing high-quality, idiomatic, modern Emacs Lisp code with incremental validation. Use when writing or modifying elisp functions, especially complex nested forms with cl-loop, multiple let* bindings, or lambdas. Covers modern features, naming conventions, documentation standards, error handling, code quality, and performance.
github-actions-troubleshooting
Troubleshoot and fix failing GitHub Actions pipeline steps. Use when user reports a failing pipeline with error output.
openspec-to-beads
Convert OpenSpec artifacts (proposal.md, design.md, tasks.md, specs/) into self-contained Beads issues for implementation tracking. Use when you have completed OpenSpec planning artifacts and need to create trackable implementation tasks, or when converting any specification or design document into actionable Beads issues.
skill-creator
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
github-actions-local-checks
Proactively run GitHub Actions pipeline checks locally before pushing code. Use when making code changes to catch test, lint, and type-check failures early.
characterization-testing
Guide for characterization testing - capturing current system behavior to enable safe refactoring. Use when working with legacy code, undocumented systems, or code that needs modification but lacks tests. Helps create a safety net before making changes by testing what the code actually does (not what it should do).
architecture-analyzer
Perform in-depth architectural analysis of codebases through iterative dialog and visual diagrams. Use when users want to understand their software architecture, explore design patterns and anti-patterns, evaluate potential architectural changes, or gain deep insights into code structure and relationships. Particularly valuable for onboarding to unfamiliar codebases, assessing technical debt, planning refactoring efforts, or understanding how components interact across system boundaries.
bead-creator
Create self-contained Beads issues for work tracking. Use when creating new beads for bugs, tasks, questions, or follow-up work - whether from specifications, discovered during implementation, or standalone requests. Handles proper formatting, labeling, dependencies, and ensures beads are actionable with embedded context.
bead-implementation
Execute on a Beads issue with proper discovery workflow and mandatory session close protocol. Use when starting work on a bead, during implementation when discovering new work that should be externalized, or when completing a session to ensure all work is properly committed.
laravel-forge
Complete Laravel Forge management via curl - servers, sites, deployments, databases, SSL, SSH keys, firewall, jobs, daemons, and more.
ace-tool
Semantic code search tool. Use `ace search "query"` for codebase search, `ace enhance "history+prompt"` for prompt enhancement. Prefer `rg` for exact matches.
rewrite
Ground-up rewrite of everything within a scope, rebuilt clean and idiomatic in the language's modern style as if written for the first time, every inherited assumption discarded - while preserving the user-facing contract so the product behaves identically. First argument is the target path (defaults to "."); anything that isn't a path is treated as extra context. Plans first (read-only, first-principles) via the plan skill and implements only after you authorize. Use when asked to rewrite, rebuild, modernize, or overhaul a file, module, or project from scratch.
plan
Research-and-plan mode for substantive tasks before implementation. Use when the user asks to plan, design, investigate, figure out, or explore how to make a non-trivial code change. Restricts the agent to read-only operations until the user explicitly authorizes implementation. Grounds every claim in the codebase and current docs instead of memory, keeps a full rewrite and rejecting the task itself on the table, and opens with the before and after that the project's users will feel.
exa
Web research using Exa search via scripts. Provides web search, advanced filtered search, code context lookup, and URL content extraction. Use when researching topics, finding documentation, looking up code examples, or extracting webpage content. No API key required.
handoff
Write or refresh a HANDOFF-<slug>.md briefing in the root of the working directory that transfers everything this session knows, so the next agent continues the work as this agent would - from that file alone, with no transcript and no re-explaining. Use when the user asks to hand off or save the session or to write down where things stand, when context is running low, at the end of a work stint, or before a long or risky operation that could end the session.
offers
Watch supermarket products and report the offers running on them. Use whenever the user wants to follow a product ("watch for an offer for Monster Energy", "tell me when X is cheap"), asks whether something is on offer or what it costs right now, or wants to see, change, or stop their watches.
obsidian
Read and write the user's Obsidian vault - notes, recipes, the journal, drafts - and keep it in step with the phone and laptop that edit it too. Use whenever they want something looked up, written down, or changed in their vault ("add this recipe", "what does my X note say", "write that down in Obsidian"), and always after you have edited anything inside it.
macros
Track daily nutrition - calories and macros (protein, fat, carbs). Use whenever the user mentions food, a meal, eating something, a food photo, calories, protein, macros, their weight, or nutrition goals.
image
Send a picture to the user on WhatsApp - a photo they asked for, an image from the chat archive, a page of a document, a chart, a screenshot. Use whenever they should look at something rather than read a description of it, or when they ask you to send or show them one.
files
Work with files - the ones the user sends you (a PDF, a zip, a spreadsheet, a video) and the ones you send back to them on WhatsApp. Use whenever a file arrives, whenever the user asks for something as a file or a download, and whenever something has to be moved out of a file into the vault or the working directory.
diagram
Draw something and send it to the user as a picture on WhatsApp - a flowchart, a sequence, a state machine, a timeline, an entity relationship, a mind map. Use when the answer is a shape rather than a sentence - how something flows, what talks to what, what happens in which order, how a set of things relate.
briefing
Send the day's briefing - one message with today's weather, daylight, calendar and offers. Use only when the user asks for the briefing itself ("send me the briefing", "brief me", "what does the briefing say"), not for the individual questions behind it.
commit
Commit with a generated Conventional Commits message, then push. Commits the staged changes; when nothing is staged, commits the whole working tree, untracked files included. Never unstages anything. Use when the user asks to commit, save work, or commit and push. The user's confirmation in the questionnaire is the explicit, single-use git approval for the staging, the commit and the push.
backup
Commit and push the working directory to its private git backup repo right now. Use when the user asks to back up, save, commit, or push the workspace, or after important changes worth persisting before the next scheduled backup.
changelog
Maintain a project's CHANGELOG.md in Keep a Changelog 1.1.0 format - record only what moved the project's user-facing surface (commands, flags, API, config, output, screens) and never internal work, pick the right category, word it for the humans who use the project, and insert it under [Unreleased] without disturbing the rest of the file. Also cuts a release section (promotes [Unreleased] to a dated version and updates the compare links) and creates CHANGELOG.md for a project that has none. Use when the user asks to add or write a changelog entry, update the changelog, record a change in CHANGELOG.md, or prepare a release section. Reads git but never commits, tags, or bumps version files.
context7
Fetch up-to-date library documentation and code examples via Context7 API. Use when you need current docs for any library/framework, to verify APIs exist, or to get version-specific code examples. No API key required.
Page 1436 of 1709 · 85428 results
