Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Testing & quality
Review a GitCode pull request from a PR number or URL when `oh-gc` must fetch PR metadata, diff, and comments, local repository code must be inspected for context, and the output should be concrete findings or a GitCode submission draft. Use for GitCode PR review or comment-submission requests that depend on `oh-gc`. Do not use for generic local review, GitHub/GitLab flows, or automatic submission without explicit user confirmation.
How to author custom PostHog Review skills: the review perspectives, blind-spot checks, validation criteria, and resolution criteria that drive PostHog Review's automated PR reviews. Use when a user wants a new review perspective (a specialist lens on their PRs), a custom blind-spot sweep, their own validation bar for which findings get published, or their own bar for which review comments get implemented. Trigger on "create a PostHog Review perspective", "custom review perspective", "my own blind-spot check", "custom validation criteria", "custom resolution criteria", "tune what PostHog Review publishes", "tune what PostHog Review implements".
Review Supabase docs changes locally in your `supabase/supabase` checkout — either an open PR (triage, classify, verify) or your own branch before opening a PR (local self-review). Covers markdown pipeline, MDX content, tutorials, examples, Studio links, and docs tooling. Use when asked to review docs PRs, self-review a draft branch, check who has approved, verify build output, or evaluate supabase/supabase documentation changes.
Verify assumptions before implementing large or risky changes to APIs, provider data, billing, pricing, quotas, or defaults. Use when a mistake could affect customers or the user asks if a change is safe to ship. Checks real-world impact beyond passing tests.
Write RSpec tests for Ruby and Rails applications. Use when creating spec files, writing test cases, or testing new features. Not for Minitest — use minitest-coder instead. Covers RSpec syntax, describe/context organization, subject/let patterns, fixtures, mocking with allow/expect, and shoulda matchers.
Run an autonomous Humanize-governed SGLang SOTA performance loop for one LLM model: first perform a fixed fair SGLang benchmark against the requested comparison framework set, then start one RLCR loop that repeatedly decides the gap, profiles the current bottleneck, runs layer/kernel pipeline analysis, patches SGLang code, optionally uses ncu-report-skill for kernel evidence, and revalidates until SGLang matches or beats the best observed requested framework under the same workload and SLA.
Signals scout for PostHog MCP tool calls. Watches `$mcp_tool_call` telemetry for tools that need improvement — broad-reach failure rates, retry hammering, slow or context-bloating responses — grouped by owning product category, each with a fix suggestion.
Audit and maintain the joelclaw skill inventory. Use when checking skill health, fixing broken symlinks, finding stale skills, or running the skill garden. Triggers: 'skill audit', 'check skills', 'stale skills', 'skill health', 'skill garden', 'broken skill', 'skill review', 'fix skills', 'garden skills', or any task involving skill inventory maintenance.
Always invoke to review, repair, or deliver geospatial or GeoAI code, including contract compliance, security, error handling, transactions, tests, scripts, functions, notebooks, packages, CI/CD, and repository changes, even when deployment is not requested. Pair with the domain skill for ETL and other production code. Covers CRS/data invariants, dependencies, cross-platform reproducibility, automation, and shipping. Do not trigger for unrelated software or analysis requesting no code or repository artifact.
Devil's Advocate stress-testing for code, architecture, PRs, and decisions. Surfaces hidden flaws through structured adversarial analysis with metacognitive depth. Use for high-stakes review, stress-testing choices, or when the user wants problems found deliberately. NOT for routine code review. Triggers on "스트레스 테스트", "stress test", "devil's advocate", "반론", "이거 괜찮아", "문제 없을까", "깊은 리뷰", "critical review", "adversarial".
Test-driven development for features, bug fixes, regressions, and safe refactors using a failing-test-first workflow. Use when Codex needs to add or change behavior with proof, reproduce a bug in a test, write regression or characterization tests, make a refactor safer, or respond to prompts like "use TDD", "red-green-refactor", "write the test first", "add a regression test", "reproduce this in a test", "prove the fix", "cover this change with tests", or "make this safe to refactor". Prefer this skill when confidence should come from executable evidence instead of reasoning alone.
Simplify and refine TYPO3 extension code for clarity, consistency, and maintainability while preserving functionality. Reviews PHP classes, Fluid templates, TCA, Services.yaml, and ext_localconf/ext_tables for TYPO3 v14 best practices. Run after implementing a feature or before merging a PR. Use when working with simplify, clean up, refine, code quality, reduce complexity, readability, maintainability, review code, polish, TYPO3 code review.
Review local changes, run or verify validation, decide commit readiness, generate a repo-native commit message, and create the git commit when explicitly requested. Use when preparing or performing a commit, checking code review findings before commit, validating changed files against a spec/plan/test evidence, preventing commits with blocking review or failed tests, or writing a commit message with intent, change, review, and validation evidence.
Audit, implement, and fix web accessibility with a screen-reader-first lens. Use when building or reviewing UI components, forms, dialogs, navigation, dynamic content, or any interactive element. Covers WCAG 2.2 AA compliance, ARIA patterns, keyboard navigation, focus management, and assistive technology compatibility (NVDA, JAWS, VoiceOver). Trigger on: "accessible", "a11y", "screen reader", "WCAG", "ARIA", or when adding any interactive UI.
Use when putting the agent in this repository on a Veris Bench bench as a "Your image" candidate, when updating that candidate to a new version, or when its bench attempts show empty replies, failed tool calls, no model calls for the grader, or never reach the bench's twins.
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
Operational rubric that turns "don't make AI slop" into observable properties, severity levels, evidence requirements, and repair actions for interface design. Use as the reference rubric when building or reviewing marketing sites, product interfaces, dashboards, portfolios, or e-commerce pages, especially alongside frontend-design.
Use when running a minimal test matrix for the Model Studio skills that exist in this repo, including image/video/audio, realtime speech, omni, visual reasoning, embedding, rerank, and edit variants. Use to execute one small request per skill and record results.
Use when benchmarking similar product documentation and API documentation across Alibaba Cloud, AWS, Azure, GCP, Tencent Cloud, Volcano Engine, and Huawei Cloud. Given one product keyword, auto-discover latest official docs/API links, score quality consistently, and output detailed prioritized improvement recommendations.
Implements RESTful API design with versioning and request specs. Use when building APIs, adding API endpoints, versioning APIs, or when user mentions REST, JSON API, or API design. WHEN NOT: Internal-only endpoints, HTML views, Turbo Stream responses, or APIs without external consumers.
Slice-based, evidence-driven framework for explicit codebase migrations and convergence programs. Use when the user is moving from a named source to a named target, running a multi-session standardization effort across a codebase, consolidating parallel implementations into one target architecture, or asking for a migration playbook with slices, ratchets, handoffs, and release closeout. Do not trigger for normal refactors, routine cleanup, or one-off architecture improvements that are not framed as a migration or convergence effort.
[Hyper] Optimize an existing codebase through baseline-first experiments, binary evaluation, and one-mutation-at-a-time iteration. Use for codebase autoresearch, measured bottleneck reduction, benchmarked code optimization, and evidence-backed refactors.
Runs one independent review of completed local changes, pull requests, branch diffs, or commits through the installed autoreview Go CLI using Codex, Claude, Cursor, or Grok; gathers authoritative acceptance criteria, selects or honors one provider, validates findings, applies scoped fixes, verifies and reruns accepted fixes, and safely reports reproducible CLI defects. Use when the user asks for an autoreview, a review of their code, pull request, or changes, an automated PR or second-model code review, a final tool-backed review, or review closeout after builder verification. Do not use as builder verification or a multi-reviewer panel.
Security testing skill for A01:2025 Broken Access Control — the #1 OWASP risk for two consecutive cycles. Use this skill whenever the user asks about: access control vulnerabilities, IDOR (insecure direct object references), authorization bugs, SSRF testing, CORS misconfiguration, privilege escalation, JWT/session manipulation, CSRF protection, or auditing FastAPI/Flask endpoints for missing auth guards. Also trigger when the user says "test my API for access control", "check authorization", "find IDOR bugs", "review permissions", "audit my routes", or mentions CWE-284, CWE-285, CWE-352, CWE-639, CWE-862, or CWE-918.
Orchestrate CAW (Cobo Agentic Wallet) agent evaluations locally on a Mac and dispatch headless claude / openclaw agents to remote servers for execution, finally producing scoring data and analysis reports. Use when: the user wants to run CAW evaluations, run an eval, test a Skill, assess agent quality, generate an evaluation report, or says "跑评测", "测评 CAW", "eval", "评分", "recipe 评测", "弱模型评测", "openclaw 评测", "模型兼容性测试".
Use when the user asks to create, update, or review a source-controlled Cekura JSON CI/CD test suite for a voice AI repository; create Tests-as-Code specs; turn a voice-agent code change into regression coverage; add deterministic Cekura voice tests to CI; or test an STT, LLM, TTS, VAD, interruption, idle-timer, DTMF, or call-lifecycle pipeline; or set up a CI gate that blocks a merge when the voice pipeline regresses. Inspects the repository before authoring a compact JSON suite, validates it safely against Cekura, and wires the workflow that runs it.
Check and fix accessibility (a11y) on front-end projects (web and mobile web), including Next.js, React, Vue, Angular. Use when the user asks about accessibility, a11y, WCAG, screen readers, voice control, Voice View, keyboard navigation, focus management, ARIA, semantic HTML, color contrast, or fixing accessibility issues in HTML, React, Next.js, Vue, or other front-end code. For native mobile apps (React Native, iOS, Android), see reference; patterns differ.
Performs a structured, comprehensive code review of git changes. Detects stack and architecture automatically, applies relevant stack/pattern-specific rules, and reports findings by severity. Use when: user runs /code-review, /review-pr, or asks to "review my changes".
Review code changes with a human in a browser instead of pasting a diff into the terminal, using the codereview-axi CLI. Use when the user wants to look at changes together, says review this with me / let's review / walk me through the diff / check my changes, or when you are about to hand over a diff large enough that they will want to annotate it rather than read it in chat.
Iterative code review and planning discussion between the local agent and Codex CLI. Orchestrates an automatic back-and-forth debate where both agents discuss findings, architecture decisions, or implementation plans until reaching consensus. Codex CLI runs READ-ONLY and never modifies files; model and reasoning effort come from the user's local Codex config. Supports plan mode: when the local agent has a plan ready, Codex evaluates and iterates on it before implementation, producing an updated consensus plan. Use when the user asks to review with codex, analyze with codex, discuss code with codex, iterate with codex, consult codex, ask codex, review the plan with codex, validate plan with codex, or any Codex CLI request for code review, architecture review, plan review, or implementation strategy. Does NOT trigger on non-code topics like diet, fitness, writing, life decisions, or general strategy; use codex-discuss for those.
Enterprise Android/Kotlin skill for 2026 — Kotlin 2.x K2 compiler, AGP 9, Navigation 3, edge-to-edge Compose, strict MVI with atomic _state.update, banned AI antipatterns. Jetpack Compose, Hilt, Room, KMP, CameraX/ML Kit, performance, testing. Use when writing Kotlin for Android, building Compose UI, MVVM/MVI architecture, debugging recomposition, Room DAOs, Navigation 3, coroutines, or asking "structure Android app", "ViewModel UiState", "collectAsStateWithLifecycle", "edge to edge", "AGP 9", "K2 compiler", "Play Store release".
Groups a git diff into review concerns and opens a local UI so humans can comprehend AI-generated code changes. Use when reviewing a PR, a branch diff, a turn diff, when a reviewed change has new commits, or when the user asks to run comprehende or to upload the report
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.