Coding standards, frameworks, refactoring, and code review (16.6K skills)
Build Expo screens that follow iOS and Android design conventions. Covers adaptive colors, icons, responsive layouts, visual effects, media, and storage, with checks for keyboard access, text scaling, and loading or saving behavior.
Generate complete UI component or screen files for a chosen framework. Use shared design tokens, named styling values, while including interaction states, keyboard access, dark mode, responsive layouts, and reduced motion support, followed by implementation checks.
Create a design project starter for Claude Code with project instructions, shared rules, design tokens, and folders for components, images, and reference screens. Add the supporting toolkit and validation dependencies, then adapt the brief and brand settings.
Improve web interface speed and stability using Core Web Vitals, measures of loading, responsiveness, and layout shifts. Review loading strategy, reserved media space, and animation behavior, then verify results under mobile network and CPU constraints.
Generate theme files for web, iOS, and Android from shared design tokens, the named values for colors and spacing. Set up platform exports, theme overrides, and automated checks that keep generated files consistent with their source.
Investigate and improve page loading, interaction responsiveness, and visual stability using real-user data and controlled browser measurements. Identify causes behind LCP, INP, and CLS issues, apply focused fixes, and repeat measurements to check the results.
Diagnose slow web loading and sluggish interactions using real user data and browser recordings. Make targeted changes to images, scripts, fonts, caching, or resource loading, then compare measurements before and after under equivalent conditions.
Run agents from local folders or git repositories using Claude, OpenAI, and other supported runners. Choose a provider, send a single prompt, inspect the system prompt, or configure authentication and repository caching.
Review code changes or complete files for security, error handling, readability, naming, input validation, and performance issues. Receive findings grouped by severity, with line references, suggested fixes, and observations about what works well.
Create an AI agent’s configuration, personality definition, and folder structure. Set model preferences, behavior rules, and runtime limits, then organize the skills, tools, and reference documents the agent will use.
Convert gitagent definitions into files for Claude Code, OpenAI Agents SDK, CrewAI, and other listed tools. Also import existing Claude, Cursor, or CrewAI configurations into gitagent files for use in another environment.
Set up gitagent and build a first agent from a template. Follow steps to configure its name, model, and personality, validate the files, run it with Claude, and share it through Git.
Explore ideas, investigate existing code, and clarify requirements before or during an OpenSpec change. Compare design options, map dependencies, and identify open questions, with optional planning documents saved within a scope you approve.
Implement pending tasks from an OpenSpec change, using its specifications and design files as context. The skill supports starting or resuming work, updates task checkboxes, verifies progress, and pauses when requirements or design issues need clarification.
Find evidence-backed opportunities to simplify codexhost, such as unused interfaces, duplicated representations, or unnecessary flexibility. Trace actual callers and documented design choices, then list proposed reductions with counterarguments and compatibility risks.
Review codexhost pull requests or local changes for concrete defects in component boundaries, asynchronous behavior, and test coverage. Produce findings that identify the affected code, explain the impact, and provide supporting evidence.
Audit and revise codexhost comments and documentation that depend on past conversations or review discussions. Preserve valid facts while rewriting explanations around the current code, so readers can understand them without access to earlier drafts or discussions.
Summarize a batch of codexhost pull requests in one table, covering purpose, scope, authors, automated checks, and review status. Assign evidence-based recommendation scores and identify which changes need closer review.
Merge an OpenSpec change’s delta specifications, which describe requirement changes, into the main specifications without archiving the change. Apply additions, edits, removals, and renames while preserving unaffected content, then validate the updated specifications.
Finalize a completed OpenSpec change by checking its documents, task counts, and specification updates. Review any outstanding work, choose whether to merge specification changes, and move the change into a dated archive.
Turn a feature or fix description into an OpenSpec change with the planning documents needed for implementation. It examines the project, clarifies significant ambiguities, and prepares proposals, requirements, designs, and tasks according to the configured workflow.
Revise an existing OpenSpec change when requirements or design decisions evolve. It compares planning documents, proposes edits to resolve contradictions and duplication, and applies confirmed revisions while leaving implementation code untouched.
Plan, implement, or review a Harness plugin that connects an execution system to codexhost. Build a capability checklist, implement supported features through public interfaces, and verify plugin loading and native behavior with evidence recorded for review.
Find relevant code by describing the behavior or functionality you need, using meaning-based search rather than exact syntax. Manage the project's search index, filter results by language or path, and inspect matching files and line ranges.