just-works

Apply when creating or editing prompts targeting Claude Fable 5 (model ID claude-fable-5). Covers effort-level selection, longer autonomous turns, brief-instruction steering, progress-claim grounding, action boundaries, parallel subagent orchestration, memory systems, early-stopping and context-budget mitigations, send-to-user tooling, reasoning-extraction refusal avoidance, and migration from Opus 4.8.

Apply when creating or editing prompts targeting Claude Opus 5 (model ID claude-opus-5). Covers response and deliverable length, progress narration, over-verification removal, subagent damping, self-correction narration, scope constraints, effort-level implications, code review harnesses, XML structure, long-context patterns, and migration from Opus 4.8. Do not apply to other Claude models — for Fable 5 use fable-5-prompting.

Ultra-compressed communication mode for conversation output. Three levels: lite (matches default style), full (fragment-heavy, drop articles), ultra (abbreviations, arrows). Default on /caveman is full. Activates when user says "caveman mode", "talk like caveman", "use caveman", or invokes /caveman. Switch with /caveman lite|full|ultra.

Apply when creating, updating, or rewriting tasks in ClickUp. Covers ClickUp-specific field discipline (due dates, priority, assignees, tags, dependencies, linked tasks, subtasks, custom fields), MCP tool discovery, space introspection, and asking-over-defaulting. Pairs with ticket-writing for body content. Project conventions always override these defaults.

Apply when writing or editing C# (.cs) files. Behavioral corrections for error handling, resource management, async patterns, data modeling, type safety, nullability, security defaults, and common antipatterns. Project conventions always override these defaults.

Apply when writing or editing Dart (.dart) files. Behavioral corrections for null safety, async patterns, type system, error handling, and common antipatterns. Project conventions always override these defaults.

Apply when implementing Domain-Driven Design patterns in Python (.py) files. Covers tactical patterns (entities, value objects, aggregates, domain events, repositories), layered architecture with dependency inversion, persistence strategies, validation boundaries, and common DDD anti-patterns. Best suited for projects with complex business rules spanning multiple entities. For top-level project layout by business capability see feature-driven-architecture-python; this skill governs the internals of slices with complex invariants.

Apply when structuring or restructuring a Python backend by business capability (vertical slices) — laying out a new project, reorganizing a layered codebase, or drawing feature boundaries. Governs top-level layout; for the internals of a complex slice see ddd-architecture-python. Covers directory layout, feature boundaries, inter-feature communication, database model ownership, migration strategies, boundary enforcement, testing across features, and migration from layered architectures. FastAPI-focused; the layout patterns port to Flask. Best suited for projects with 5+ distinct features.

Apply when writing or editing Flutter widgets in Dart (.dart) files. Behavioral corrections for widget composition, BuildContext, state management, performance, lifecycle, and common antipatterns. Pairs with dart-coding for language rules. Project conventions always override these defaults.

Apply when creating or editing prompts targeting the Gemini 3 family (3.0, 3.1 Pro, 3 Flash, 3.1 Flash-Lite). Covers prompt layering (system instruction + context + task), thinking_level tuning, structured output prompting, function calling guidance, grounding prompt wording, few-shot examples, long-context patterns, persona/constraint alignment, prompt decomposition, agentic workflows, and migration from Gemini 2.5.

Apply when creating or editing prompts targeting GPT-5.5. Covers outcome-first prompting, reasoning-effort calibration, verbosity tuning, long-context handling, ambiguity management, tool persistence and validation, preamble and phase patterns, retrieval budgets, structured extraction, citation discipline, personality and collaboration style, coding autonomy, frontend prompting, and migration from GPT-5.4, GPT-5.3-Codex, or older GPT models.

Apply when writing or editing PlantUML (.puml, .plantuml, .pu) files or when generating diagrams from text descriptions. Covers diagram type selection, syntax conventions, modern styling, preprocessing, audience-aware abstraction levels, and common anti-patterns. Project conventions always override these defaults.

Apply when writing or editing Python (.py) files. Behavioral corrections for error handling, resource management, async patterns, data modeling, type safety, security defaults, and common antipatterns. Project conventions always override these defaults.

Apply when writing or editing React (.tsx/.jsx) files. Behavioral corrections for hooks, state management, component structure, memoization, security, and common antipatterns. Project conventions always override these defaults.

Apply when designing or implementing REST API endpoints, routes, or controllers. Covers URL conventions, HTTP methods, status codes, error responses, pagination, versioning, authentication, security, caching, file uploads, health checks, and common API antipatterns. Framework-agnostic HTTP-level patterns. Project conventions always override these defaults.

Apply when generating shadcn/ui code — signaled by components.json in the project root and components under components/ui/. Covers the copy-paste component model, Radix UI compound components, theming with CSS variables, Form/DataTable integrations, and CLI conventions. Does NOT cover general React or Tailwind CSS patterns.

Apply when the user explicitly asks to estimate a task ("how many points", "estimate this") or when a ticket-creation flow explicitly requests story points. Provides a calibration framework for assigning accurate Fibonacci story points (1–13) based on work volume, complexity, uncertainty, and risk. Triggers context gathering (code exploration, clarifying questions) before committing to a point value. Pairs with ticket-writing for body content and clickup-tickets for field discipline. Project conventions always override these defaults.

Apply when writing or editing Swift (.swift) files. Behavioral corrections for error handling, concurrency, memory management, type safety, protocol-oriented design, security defaults, and common antipatterns. Project conventions always override these defaults.

Apply when writing or editing Tailwind CSS classes in any template or component file. Behavioral corrections for dynamic styling, class composition, responsive design, dark mode, interaction states, accessibility, and common antipatterns. Project conventions always override these defaults.

Apply when writing, rewriting, or reviewing issue-tracker tickets (Jira, Linear, ClickUp, GitHub Issues, Asana). Covers titles, sections per type (PBI / Bug / Spike / Discovery), acceptance-criteria formats, bug reproduction fields, INVEST, linking, and common anti-patterns. Tool-agnostic. Project conventions always override these defaults.

Apply when writing or editing TypeScript (.ts and .tsx) files — language-level rules; React-specific patterns live in react-coding. Behavioral corrections for error handling, async patterns, type system, module system, security defaults, and common antipatterns. Project conventions always override these defaults.

Forces the minimal working solution — least code, fewest files, no speculative abstraction. Channels a senior dev applying YAGNI: stdlib, native platform features, and existing dependencies over custom code. Three levels: lite (name the lazier option, user picks), full (enforce the ladder), ultra (deletion-first extremist). Activates when user says "be lazy", "yagni", "simplest solution", "minimal", "over-engineered", "too much code", or invokes /minimal-coding. Governs WHAT you build, not HOW you talk.