Self-improvement loop triggered after any user correction. Captures mistakes as prevention rules in tasks/lessons.md so the same error is never repeated.

Review animation.c for C issues — memory safety, terminal handling, signal safety, and game logic

Audit a restaurant's web presence — check what AI models see when deciding whether to recommend it. Covers website, Google Business Profile, review platforms, and structured data.

Draft a detailed Non-Disclosure Agreement between two parties covering information types, jurisdiction, and clauses needing legal review. Use when creating confidentiality agreements or preparing an NDA for a partnership.

Review authentication and authorization design including OAuth, JWT, token expiration, RBAC/ABAC, and privilege escalation risks.

배포 스크립트의 보안·안전성 리뷰를 수행한다. 롤백 절차, 장애 대응, 권한 설정을 점검한다.

Confidence-based model tier escalation for agent tasks. Defines when to escalate from Haiku to Sonnet to Opus, escalation signals, and the never-skip-tiers rule. Use when an agent fails or underperforms.

Writes React components without unnecessary useEffect. Use when creating/reviewing React components, refactoring effects, or when code uses useEffect to transform data or handle events.

Post-draft diagnostic for sermon and book writing. Scans for machine tells, assesses authenticity risk, checks voice continuity against the author's sermon corpus. Fires after writing is complete or when reviewing existing drafts. For during-writing standards, see like-a-human.

Triage and review Linear issues - list, filter, prioritize, and update issue status. Use when asked to check tickets, review backlog, or see what needs doing.

Meta-level analysis of plan vs execution to identify process improvements. Analyzes divergence patterns and suggests updates to CLAUDE.md, agents, and skills.

PANaCEa design system reference — typography tokens, visual hierarchy tiers, shared component helpers, and inline style conventions. Use this skill whenever modifying, creating, or reviewing UI components in the PANaCEa clinical library, reference views, toolkit, or any component using inline styles. Also trigger when the user mentions typography, fonts, styling, visual hierarchy, design tokens, color system, dark mode, or consistent look-and-feel across PANaCEa components.

Load this skill when writing tests, reviewing test coverage, setting up testing strategies, or ensuring code quality. Covers the testing pyramid, unit/integration/e2e test standards, naming conventions, and mandatory 90% coverage requirements.

startproject

Run a weekly retrospective across Threads and note outputs, then propose next-step improvements. Use for recurring optimization.

Product management: PRDs, RICE prioritization, metrics

Review code for design token violations, SCSS convention drift, and theme wrapper compliance. Compares source files against pf-tokens-SSOT.json, MUI-default-theme-object.json, and the project's styling rules. Use when the user asks to review code, audit styling, check token usage, or review a PR for convention compliance.

Load the shared GO execution workflow from `.agent/skills/go/shared/go.md`.

METAINFORMANT rules for directory scripts/menu. Use when editing, adding tests, or reviewing code under this path. Read the linked AGENTS.md first; use uv only, write outputs to output/, no mocks.

**모두**​이(가) true인 경우에만 적용:

Reviews changes for security, scalability, performance, code reuse, and maintainability using a structured checklist. Use when the user asks for a code review, wants feedback on a PR, or wants to harden code quality.

Review React/Next.js components, server actions, and data-fetching logic in the ClipRace repo to prevent performance regressions by checking RSC vs client boundaries, bundle size, waterfalls, CLS, and large list rendering. Use when the user asks for a performance review of specific routes, components, or flows.

METAINFORMANT rules for directory src/metainformant/multiomics. Use when editing, adding tests, or reviewing code under this path. Read the linked AGENTS.md first; use uv only, write outputs to output/, no mocks.

Ruby Style Guide (rubystyle.guide) conventions. Use when writing, formatting, or reviewing Ruby code for layout, naming, flow of control, methods, classes, and idioms. Complements RuboCop.

Create a six-frame storyboard that shows a user's journey from problem to solution. Use when you need a fast narrative for alignment, concept reviews, or demos.

Use when resolving random JIRA issues. Fixes bugs, refactors code, performs code and security reviews, ensures 100% test coverage, runs CI checks, and creates pull requests. Links PRs to JIRA issues and updates issue status.

Reviews and applies lightweight UI motion patterns with accessibility-safe behavior. Use when adding animations, hover/tap interactions, or section reveal effects.

Triage and review all open GitHub issues and PRs against project templates and contribution guidelines

METAINFORMANT rules for directory config/amalgkit. Use when editing, adding tests, or reviewing code under this path. Read the linked AGENTS.md first; use uv only, write outputs to output/, no mocks.

METAINFORMANT rules for directory scripts/simulation. Use when editing, adding tests, or reviewing code under this path. Read the linked AGENTS.md first; use uv only, write outputs to output/, no mocks.

METAINFORMANT rules for directory scripts/gwas/structure. Use when editing, adding tests, or reviewing code under this path. Read the linked AGENTS.md first; use uv only, write outputs to output/, no mocks.

Single source of truth for iOS-native mobile web UX. Covers viewport units, safe areas, Dialog-to-Drawer conversion, Tabs-to-vertical-stacking, touch targets, input zoom prevention, and responsive patterns. Use when building any UI component, fixing mobile issues, reviewing responsive code, or when mobile UX is mentioned.

Reviews a pull request from a PR URL by directly fetching the URL content (no `gh` dependency) and verifies compliance with CONTRIBUTING.md. Use when the user asks for a PR review, to review changes before merge, or to check contribution guidelines.

Evaluate novelty and contribution clarity (are contributions stated and supported?).

Performs a final quality pass fixing alignment, spacing, consistency, and micro-detail issues before shipping. Use when the user mentions polish, finishing touches, pre-launch review, something looks off, or wants to go from good to great.

Defines usecase-layer guidance for this repository in DDD x Clean Architecture, including usecase responsibilities, orchestration, transaction boundaries, DTOs, repository and external service ports, `internal/usecase` splitting, and `docs/domain/usecase` maintenance. Expected outputs include concise usecase-layer recommendations with reasons, placement guidance, transaction and side-effect handling notes, or small Go-oriented usecase templates with assumptions. Use when designing, implementing, reviewing, or documenting the usecase layer in this repository.

Audit email2telegram release artifacts for secret leaks and insecure defaults on GitHub, Docker Hub, and VPS. Use when preparing a release, reviewing .env/Dockerfile/compose/deploy scripts, changing scripts/release.sh, or when the user mentions security, secrets, leak, audit, hardening.

Open a GitHub pull request for this repository using gh CLI, including branch checks, push, a reviewer-friendly PR body, and verified GitHub issue links (Closes/Fixes/Refs). Always resolve or create related issues before gh pr create so the PR can link to tickets; when drafting new issues, follow the create-github-issue skill for title and body structure. **PR title and body prose are humanized** per the humanizer skill (`.cursor/skills/humanizer/SKILL.md`). Use when the user asks to create/open/submit a PR or pull request, and format title and description with the pr-title-description skill structure (including Related issues).

QA Engineer role that audits the Jasmine kindergarten website for performance (Core Web Vitals), accessibility (WCAG 2.2 AA), SEO, security, browser compatibility, mobile optimization, and analytics. Use when the user asks to review, audit, test, check quality, run a QA checklist, fix performance, improve accessibility, or optimize the site.

Performs a structured security review with a comprehensive checklist (secrets, input validation, injection, authz/authn, XSS, CSRF, rate limiting, sensitive data exposure, dependency hygiene). Use when adding authentication, handling user input/uploads, creating API endpoints, working with secrets/credentials, integrating third-party APIs, or implementing payment/sensitive features.

Use when writing new code, renaming files, adding imports, creating error classes, or reviewing code style. Covers file naming, export naming, import patterns, validation helpers, and error handling conventions for this project.

>-

Design engineering principles and patterns for building polished, accessible web interfaces. Use this skill when building UI components, reviewing frontend code, implementing forms, handling touch interactions, optimizing performance, or creating marketing pages. Triggers on: design engineering, UI polish, input fields, form validation, button states, touch devices, mobile UX, accessibility, a11y, keyboard navigation, aria labels, font rendering, typography, layout shift, z-index, animations, transitions, easing, hover effects, tap targets, iOS Safari, prefers-reduced-motion, marketing pages, landing pages, dark mode, theme switching, scrollbars, gradients, shadows, virtualization, preloading.

'Analyze Datadog error logs for Packmind production services (API, MCP server, Frontend), group them into patterns, perform root cause analysis against the codebase, and produce a structured bug report. This skill should be used when investigating production errors, triaging bugs, auditing service health, or performing periodic error reviews. Also triggers when the user mentions Datadog, production logs, error analysis, prod issues, service health, or asks about what errors are happening in prod. Also triggers on references to specific Datadog service names like api-proprietary, mcp-proprietary, or frontend-proprietary.'

Create a pull request from the current branch to a target branch with proper formatting, Jira link, and reviewers

Perform a security pass aligned with MediaJira standards (no secrets, auth/CORS, input validation, least privilege, credential handling). Use for pre-release checks or when touching auth/sensitive code.

Scans TypeScript/JavaScript source files to find files that still contain hardcoded, non-i18n text and outputs only the file paths. Use when auditing localization coverage, checking a directory, or reviewing a commit for missing i18n extraction.

Applies official Google best practices when writing or editing Gemini system prompts (systemInstruction). Use when creating or changing system prompts for Gemini (e.g. transcription, Dictate Prompt, Prompt & Read), when reviewing prompt text in AppConstants or SpeechService, or when the user asks about Gemini prompt design.

Verifies that i18n replacements (t/i18n.t/Trans) in source code follow extract-i18n-keys rules and that keys exist in modules with correct English copy; also checks for over-extraction (content that should not have been extracted, e.g. URLs, class names, alt/aria-label/data-*, meta, technical strings). When given a commit or range, scopes verification to changed files and optionally checks that replaced text matches module values via diff. Use for PR/commit review, i18n audit, or post-replacement validation. Read-only.

Запускает трёх суб-агентов: **task-decomposer → task-worker(s) → work-reviewer**. Каждый шаг выполняется **в отдельном агенте** (отдельном контексте), а не в одном общем.