ryanthedev

Decomposes underspecified requests by classifying gaps (missing info, ambiguity, false premises) and generating targeted clarifying questions. Produces a confirmed problem statement before any implementation begins. For sharpening a request that already exists; not for open-ended requirement discovery from a vague idea (use the research command).

Implements the Standard/Full planning pipeline for Medium and Complex tasks: multi-step discovery, phase decomposition with skill matching, cross-cutting concerns, and plan emission with Gate fields.

Placeholder skill directory required by skill-eval's run_eval tool signature. This benchmark runs configurations=['without_skill'] only — we are measuring raw model tiers, not a skill — so this file's content is never mounted or read by a subject session.

Guides module and API design using APOSD principles: generates multiple design alternatives, compares them on information hiding and interface depth, and produces a documented design decision. For creating new module/API design; not for assessing existing designs (use aposd-reviewing-module-design) or routine-level design (use cc-routine-and-class-design).

Assesses existing module and interface design for complexity symptoms: information leakage, shallow interfaces, pass-through layers, and unknown unknowns. Produces a structured assessment — not transformations (use aposd-simplifying-complexity to edit) and not new-design generation (use aposd-designing-deep-modules).

Transforms complex code by applying APOSD's pull-complexity-downward principle: resolves error hierarchies, collapses configuration, and moves caller-side logic into modules. Produces edited code; not for assessment-only review (use aposd-reviewing-module-design) or designing new modules from scratch (use aposd-designing-deep-modules).

Verifies implementation completeness across functional correctness, error handling, concurrency, and security dimensions using APOSD's post-implementation checklist. Run after a coding task is nominally complete, not during active bug investigation (use cc-debugging).

Audits and restructures control flow using Code Complete's nesting, cyclomatic complexity, loop design, guard-clause, and boolean simplification rules. Covers deep nesting, loop exit design, index naming, and callback chains. For statement-level flow inside a routine; not for routine/class structure (use cc-routine-and-class-design) or staging a broader refactor (use cc-refactoring-guidance).

Subagent dispatch guidance for the Agent tool — whether to delegate or work inline, the four-part delegation contract, model and effort selection, parallel fan-out sizing, fork vs fresh subagent, and dispatching independent verifier agents. Use before any Agent tool call: spawning subagents, fanning out work across files or research angles, delegating exploration or review, briefing a verifier, or deciding whether delegation is worth it at all. Not for: wording a prompt that isn't an agent dispatch (use oberskills:prompt) or authoring reusable skill and agent definition files — structure, frontmatter, and evals (use oberskills:skill-craft); their prompt bodies (use oberskills:prompt).

Controls a live Chrome browser through a persistent puppeteer-core MCP connection — click, type, hover, drag, fill forms; navigate pages; take screenshots and export PDFs; read the DOM and accessibility tree; run Lighthouse audits and Core Web Vitals traces; intercept, record, stub, or block network requests; export HAR traffic; emulate devices and network conditions; manage browser storage; upload and download files. Use when interacting with or automating a running browser, clicking elements, taking screenshots, running Lighthouse, capturing network traffic as HAR, checking the accessibility tree, intercepting or stubbing requests, or emulating a device or viewport. Not for: web search or fetching public URLs without a browser session (use web-research), reading a static HTML file already on disk, fixing TypeScript or JavaScript build or compile errors, debugging source code without a running page, or tasks that don't require controlling a live browser.

Design and review prompts Claude-first — system prompts, reusable agent definitions, long or novel dispatch briefs, pipeline stages, and prompts users will run elsewhere. Two modes: design and adversarial review. Covers Claude model behavior and migration, de-prompting, few-shot design, output-schema ordering, context engineering, prompt security, optimization, non-Claude/Codex target review via references/porting.md in this skill directory, and a verbatim behavior-snippet library. Use when writing, improving, debugging, reviewing, or migrating any prompt. Not for: creating, reviewing, or testing skill files (use skill-craft); routine subagent delegation prompts (use agent).

Captures and summarizes screenshots from the local desktop or a named window using the bundled capture script. Use when the user asks for a screenshot, screen capture, active-window capture, visual inspection of the current screen, or named-window screenshot. Not for browser-page screenshots when browser MCP tools are already controlling a page.

Create, evaluate, and review Claude Code skills and reusable agent definition files (structure, frontmatter, evals). Covers the skill-vs-subagent-vs-hook decision, SKILL.md authoring and frontmatter, trigger-description optimization, baseline-first evals with pressure testing via the skill-eval MCP tools, validation, and packaging. Use when creating a new skill, improving or benchmarking an existing one, reviewing a skill directory, or writing evals for a skill. Not for prompt wording design or review — including the prompt body inside an agent definition (use oberskills:prompt), live subagent dispatch (use oberskills:agent), or prose editing (use oberskills:write).

Runs parallel multi-angle web research at four depth modes (scan, brief, breadth, deep): a planner grounds the query in local context and plans search dimensions, parallel search agents pull verbatim extracts with source URLs, one dimension always hunts counter-evidence, and synthesis is grounded back in the local project. Use when researching a topic, library, or tool on the web, comparing options or products, surveying a landscape of approaches, or checking current, version-sensitive, or time-sensitive information. Not for: driving or automating a live browser session (use oberskills:browser), academic literature search across arXiv or Semantic Scholar, or questions about the current repo, project, or files already on disk — read those directly.

Edits and reviews prose to remove the AI tells a model cannot see in its own writing — measured detection signals, not intuition: structural rhythm first (sentence-length variance, information density, discourse order), the surface kill-list second (banned phrases, em-dash overuse, hollow openers, hedging), with an optional phased interactive review for longer pieces. Use whenever prose for human readers is being written or edited — blog posts, announcements, docs, emails, READMEs — when text needs humanizing or sounds robotic, and when reviewing someone's writing. Not for: wording prompts or system-prompt text (use oberskills:prompt), converting markdown for platforms like Slack or Notion (use penman), or writing code, comments, docstrings, or API reference docs.

Decomposes user intent through structured brainstorming before acting on ambiguous or underspecified requests — classifies the ambiguity type, generates competing interpretations, and selects the clarifying question that removes the most uncertainty. Use when requirements are unclear, a request could have multiple valid interpretations, critical details or constraints are missing, or the request contains contradictions. Most requests need no clarification — invoke only when a concrete ambiguity would change the approach. Not for: mechanical or well-specified tasks (renames, typo fixes, version bumps, running tests or commands, any edit with an explicit file and target), details you can determine yourself from the codebase, or full research or planning workflows that own their own scoping.

React Native coding guidance — consults official docs before writing code and suggests verification after. Use when writing components, implementing features, building screens, fixing queries, or adding views. Triggers on "write a component", "implement this feature", "build this screen", "add a view", "React Native code", "fix this query", "use best practices".

Audit accessibility of the running React Native app. Captures the accessibility tree from the iOS Simulator and checks it against React Native best practices. Triggers on "audit accessibility", "check a11y", "VoiceOver check", "accessibility issues", "screen reader test", "are labels correct".

Intelligent environment reset for React Native projects — diagnoses stale caches before cleaning instead of blindly nuking everything. Triggers on "clean build", "reset caches", "fresh start", "metro won't start", "build is broken", "nuclear clean", "clear derived data".

Debug React Native apps — view console logs, JS errors, Metro status, evaluate expressions, inspect React component tree, monitor network requests. Use for runtime debugging of JavaScript and React layers. Triggers on "console logs", "JS errors", "Metro status", "evaluate expression", "React component tree", "network requests", "debug".

Test deep links and URL schemes in the iOS Simulator. Reads navigation/linking config, constructs test URLs, fires them via xcrun simctl, and verifies the resulting screen. Triggers on "test deep link", "test this URL scheme", "verify navigation to", "does deeplink work", "open URL in simulator", "test universal link".

Diagnose React Native errors by cross-referencing error text against docs, error pattern database, and project config. Use for build failures, red screens, runtime crashes, and dependency issues. Triggers on "diagnose this error", "why is my build failing", "what does this red screen mean", "Unable to resolve module", "null is not an object", "Invariant Violation", "pod install failed", "gradlew failed", "EACCES", "metro error".

Search React Native official documentation. Use when answering questions about React Native APIs, components, hooks, layout, navigation, native modules, or platform-specific behavior. Triggers on "React Native", "RN docs", "how does View work", "FlatList props", "native module", "new architecture".

Control the iOS Simulator — tap, type, swipe, screenshot, record video, install and launch apps. Use when interacting with the iOS simulator during React Native development. Triggers on "simulator", "tap", "swipe", "screenshot of simulator", "install app", "launch app", "what's on screen", "record video", "accessibility tree", "tap by label", "tap by id", "list elements", "go back", "scroll to top", "scroll to bottom".

Decomposes underspecified design requests by classifying gaps (missing brand context, visual ambiguity, false premises, scope faults) and generating targeted clarifying questions. Produces a confirmed design brief before any design work begins. For sharpening a design request that already exists; not for open-ended design brief discovery from a vague idea (use /design-for-ai:research).

Applies data visualization and information design principles to encode data truthfully and efficiently. Use when the question is about charts or graphs — chart type selection (bar, line, scatter, heatmap, small multiples), data-ink ratio and chartjunk reduction (Tufte VDQI), preattentive attributes (Knaflic Storytelling with Data), dashboard layout and KPI design (Few Information Dashboard Design), truthful encoding and chart lie detection (Cairo How Charts Lie), marks and channels (Munzner), colorblind-safe palette selection for charts, or chart accessibility. Not for: brand or UI color palette generation (use core color mode); overall page composition and visual hierarchy (use core audit mode); data-display tables as a UI pattern (use usability); typeface or font selection (use core fonts mode); UI copy or writing (use content-design).

Produces self-contained HTML/CSS mockups and wireframes from design tokens and page specs — the output is always a viewable .html file, not guidance or code. Use when the request is to see a design, make a mock, build or show a prototype, create a wireframe, render a preview, or visualize what a page looks like before writing production code. Reads DESIGN.md tokens and JOURNEY.md page specs when present; uses greyscale wireframe defaults when absent. Renders via the browser MCP and returns a screenshot path when connected; emits the .html with an open note when not. Not for: choosing or defining fonts, colors, visual identity, or design tokens (use core design); critiquing, auditing, or reviewing an existing design (use core audit); planning IA, flows, or page specs (use journey); building a design token system or component library (use design-systems).

Adjudicate whether users can operate an interface, and run heuristic evaluation. Use when the question is operability rather than looks — is it usable, hard to use, confusing, where do users get stuck; heuristic evaluation, Nielsen heuristics, usability audit, severity rating; the UX laws (Fitts, Hick, Miller/Cowan, Jakob, Tesler, Doherty, peak-end, Zeigarnik, aesthetic-usability); Gestalt grouping, affordances, signifiers, mapping, feedback, cognitive load, recognition over recall; picking among navigation, form, search/filter, table/data, feedback, action, disclosure, or onboarding patterns; or usability-side accessibility (WCAG POUR, inclusive design). Not for: the visual look or aesthetic audit (use core design/audit); the words/microcopy themselves (use content-design); the route through time, IA, or funnel (use journey); persuasion or conversion mechanics (use behavioral); truthful chart/data encoding (use data-viz).

Guides distributed system design using a structured 4-step framework: scope, high-level design, deep dive, wrap-up. Two modes: CHECKER audits existing designs for missing components, wrong tradeoffs, and anti-patterns. APPLIER walks through designing systems step-by-step with clarifying questions, component selection, and scaling strategies. Covers 15 system designs (rate limiter, consistent hashing, key-value store, unique ID generator, URL shortener, web crawler, notification system, news feed, chat system, search autocomplete, YouTube, Google Drive, scaling foundations, estimation, interview framework) plus cross-cutting techniques and review checklists. Triggers on: system design, distributed system design, design a system, design a distributed, scale system, architecture review, high-level architecture, back of envelope, capacity planning, design with load balancer, database sharding strategy, caching strategy for scale, CDN architecture, message queue design, microservices system design, API system desi

Start or continue threaded conversations between Claude instances across machines using grug-brain. Use when the user wants to coordinate between Claude sessions, send a message to another Claude, discuss something cross-machine, open a conversation thread, reply to another agent, check conversation status, or when troubleshooting involves back-and-forth between two computers. Also triggers on 'talk to the other Claude', 'start a thread', 'conversation with', 'cross-machine', 'reply to hive', 'check the conversation', 'post a message', 'coordinate between sessions'.

Enumerate all tmux sessions and windows via the claude-mux MCP, AI-summarize each window's state, and atomically write the grid-notify dashboard's JSON status file at $XDG_STATE_HOME/thegrid/tmux-status.json. Use when asked to refresh the tmux dashboard, update or regenerate the tmux status file, or write a status snapshot of all tmux sessions. Not for: listing, attaching to, creating, killing, renaming, or sending keys/commands to tmux sessions or windows — those are direct claude-mux tmux operations, not a status-file write.