Claude Skills by curiositech

Audit or design a status/health surface so it does not share fate with the system it describes and cannot quietly attest to itself: split read plane (independently deployed reporter + dead-man switch), three-valued verdicts (healthy/degraded/unknown — never green-by-default), external anchoring of any "tamper-evident" status chain, and the availability-inversion check (justify monitoring dependencies per capability). Use when adding a status page, deep-health endpoint, reachability verdict, o...

Reconstructs the strongest version of a position that remains attributable to its holder or canonical sources, then binds every premise to provenance, uncertainty, counterevidence, and a falsifier before critique. Use for fair reciprocal review, disagreement mapping, or decision preparation. NOT for persuasive optimization of abuse, discrimination, atrocity, scams, coercion, deceptive influence, mind-reading an absent person, crisis support, clinical interpretation, or treating a reconstructi...

A graph-based shared blackboard where agents coordinate indirectly by depositing typed pheromone traces (PHEROMONE, BELIEF, PREFERENCE, ANTIBODY, RESOLUTION) onto graph nodes, then sensing and following concentration gradients. Traces spread via a stated scalar graph update and decay, creating a lossy attention field. It can prioritize inspection; it does not replace reliable messaging, assignment, delivery, coverage, or completion evidence. NOT for treating a decayed attention trace as durab...

Use when designing log schemas, choosing JSON vs text logs, setting up correlation IDs across services, redacting PII, controlling log volume costs, or wiring logs into OpenTelemetry. Triggers: "logs are unsearchable", trace_id missing from logs, PII leak in production logs, log volume bill spike, log levels misused, structured fields vs string interpolation, parent_span_id propagation, sampled vs always-log decisions, log routing (vendor + cold storage). NOT for log aggregation tooling speci...

Supabase administration, RLS policies, migrations, and schema design. Use for database architecture, Row Level Security, performance tuning, auth integration. Activate on "Supabase", "RLS", "migration", "policy", "schema", "auth.uid()". NOT for Supabase Auth UI configuration (use dashboard), edge functions (use cloudflare-worker-dev), or general SQL without Supabase context.

Defend long-running ML/AI daemons against supply-chain attacks delivered via third-party extension ecosystems — ComfyUI custom nodes, VS Code extensions, npm postinstall, Hugging Face Spaces, Ray clusters, Triton/Cog plugins, browser extensions. Activate on: supply chain attack, malicious package, custom node security, dependency confusion, package backdoor, cryptominer in dependency, Akira / Pickai / leftpad / event-stream / ua-parser-js, Sigstore signing, lockfile pinning, capability sandbo...

Designs and audits controller-owned multi-agent invocation contracts: capacity reservation, single-use leases, generation fences, idempotent dispatch, typed gathers, deterministic reduction, bounded rework, cancellation, settlement, and effect-denied replay. Use when projecting an approved execution plan onto bounded concurrent work. NOT for task decomposition, choosing a planning topology, context partitioning, IPC or framework selection, dialogue design, live agent operation, or granting ru...

Techniques for surfacing implicit, hard-to-articulate knowledge that experts use in practice

Use when migrating Tailwind v3 to v4, configuring CSS-first @theme tokens, debugging Oxide engine errors, dealing with the absence of tailwind.config.js, container queries (@container) usage, layer cascade problems, or "dynamic class names not generating". Triggers: switching from `@tailwind base` to `@import "tailwindcss"`, RSC/Next.js streaming + CSS ordering, dark-mode strategy choices, custom variants via @variant, content-scanning glob tuning. NOT for Tailwind v3 (different config model)...

Analysis of how suppressing volatility creates fragile systems vulnerable to black swan events

Designing TanStack Query (React Query) v5 for server-state management — query key factories, staleTime vs gcTime, partial-match invalidation, optimistic updates with rollback, prefetching, and Next.js App Router hydration. Grounded in tanstack.com v5 docs. NOT for pure client state (useState/Zustand), form state, GraphQL cache design, or true real-time push (WebSockets/SSE).

Reconstruct and review Grassi's bounded simulator task-analysis procedure: GOMS-like goals/methods/selection rules, critical cue inventories, and five-participant CDM review. NOT for live ship handling, operational navigation, or general claims about expert cognition.

Turn formally verified infrastructure into developer adoption with talks, blog posts, onboarding docs, venue strategy, and community building for TLA+, model checking, mechanism design, and collateralized contracts. Use when audience, depth, medium, and call-to-action must match technical adoption goals. NOT for academic paper writing, non-technical marketing, social media management, or PR.

Use when designing Terraform modules for reuse, structuring root vs child modules, choosing variable shapes (object vs flat list), wiring versioning + module registries, managing remote state, detecting and remediating drift, or refactoring monolithic configs. Triggers: module composition, var validation blocks, output contracts, count vs for_each, lifecycle ignore_changes, terraform import, state mv between resources, workspaces vs separate state files, OIDC to provider, drift detection runs...

How the best math/CS textbooks (Halmos, Axler, SICP, Knuth, Feynman, Sipser, CLRS, MacKay, Pólya, Lakatos, Hersh) open a chapter, place worked examples and exercises, and state honest boundaries — applied to the Harbor Book's chapters. Use when drafting or reviewing a chapter opener, deciding where an exercise or worked example belongs, auditing a chapter's page grammar or claim-labeling, or running scripts/chapter_lint.py against a .tex chapter. NOT for the seven-moves result-writeup style (...

license: Apache-2.0 NOT for unrelated tasks outside this domain.

license: Apache-2.0 NOT for unrelated tasks outside this domain.

Use for rapport, elicitation, transcript audit, interview design, and disclosure-yield analysis by auditing attention, coordination, and typed yield instead of vague warmth or self-report. NOT for coercive extraction, generic tone polishing, or persuasion tasks that do not depend on genuine disclosure.

license: Apache-2.0 NOT for unrelated tasks outside this domain.

license: Apache-2.0 NOT for unrelated tasks outside this domain.

license: Apache-2.0 NOT for unrelated tasks outside this domain.

license: Apache-2.0 NOT for unrelated tasks outside this domain.

Design and review an explicitly modelled multi-party agent-labor market, including participation groups, cross-group effects, information and control boundaries, and settlement outcomes. NOT for payment plumbing, a pricing theorem, or an unverified escrow operation.

Diagnose the information structure behind a claim, choose a truthful publication-grade visual grammar, and author, render, and audit compact TikZ/PGFPlots figures. Use for LaTeX whitepaper figures, technical diagrams, quantitative plots, protocol explanations, and figure redesigns when semantic fit, layout, or legibility is uncertain. NOT for decorative cover art, image generation, data extraction, or generic slide diagrams.

You MUST use this skill when writing, running, or verifying TLA+ specifications for distributed systems, coordination protocols, or state machine designs. This skill is opinionated and prescriptive — it will push you toward bounded model checking with TLC, concrete worked examples, and measurable quality gates. NOT FOR: general formal methods theory, Coq/Lean/Isabelle proof assistants, SPIN/Promela model checking, or theorem proving without explicit state machines.

Structure and analyze arguments using the Toulmin model (Claim-Data-Warrant-Backing-Qualifier-Rebuttal). Use when the user wants to build logical arguments, evaluate reasoning, prepare formal debate, clarify messy reasoning, write persuasive pieces, or mentions 'argument structure', 'warrant', 'backing', 'logical framework', 'structure this argument'. Useful as input for speech-writing, fallacy detection, and steelmanning.

Use for cognitive task analysis, CTA method selection, knowledge capture, tacit expertise routing, and representation design by matching elicitation methods to knowledge type. NOT for generic ontology naming, benchmark-only model evaluation, or ML architecture tuning.

Use for cognitive task analysis, CTA method selection, knowledge capture, tacit expertise routing, and representation design by matching elicitation methods to knowledge type. NOT for generic ontology naming, benchmark-only model evaluation, or ML architecture tuning.

Use when integrating Hugging Face Transformers.js (`@huggingface/transformers`) for in-browser or in-Node inference, debugging model and dtype loading, building bi-encoder / cross-encoder / classification pipelines, configuring model cache directories, or bypassing high-level pipelines to read raw logits. Triggers: "model failed to load", cross-encoder scores all 1.0 (softmax-over-1 trap), env.allowLocalModels confusion, cacheDir overrides, ONNX runtime mismatch, ESM vs CJS pipeline imports, ...

Designs and audits deterministic evaluation envelopes for agent-control systems using sealed schedules, typed faults, independent raw evidence, oracle mutation, semantic replay, counterexample minimization, and result vectors. Use for simulation, model, replay, shadow, or canary claims. NOT for launching a runtime, authorizing a canary, certifying containment from a model, self-witnessing success, or treating static usability fixtures as live accessibility proof.

Compiles provenance-bound, authority-filtered Context IR for continuation or body translation, including obligations, omissions, redactions, retrieval-space identity, and directive eligibility. Use before preparing a successor or cross-backend handoff. NOT for admitting a body, fencing a predecessor, granting tools, treating memory as authority, retrieving across incompatible vector spaces, or copying raw secrets.

Applies Edward Tufte's evidence-design doctrine — data-ink ratio, graphical integrity and the Lie Factor, small multiples, sparklines, layering and micro/macro readings, the fundamental principles of analytical design, and sentences-over-bullets — to choose and audit how a claim should be shown: a table, a sparkline, a small-multiples grid, an annotated graphic, a margin figure, or a plain sentence. Use when designing or reviewing a chart, table, dashboard, figure, slide, or margin apparatus ...

Set up Cloudflare Turnstile end-to-end in a project. Scan the codebase, create the widget via the Cloudflare API, embed it on the right forms, wire canonical server-side siteverify in the customer's existing backend, validate, and persist the skill. Load this when a user asks to add Turnstile, set up CAPTCHA, protect a form from bots, or fix a Turnstile integration. Mirrors developers.cloudflare.com/turnstile/spin.

Advanced TypeScript type system patterns for production codebases. [What: branded types for nominal typing, discriminated unions, template literal types, conditional types, the infer keyword, satisfies operator, const assertions, Zod schema inference, type-safe event emitters, exhaustive switch checking] [When: designing domain models, building type-safe APIs, creating reusable generic utilities, eliminating runtime bugs with compile-time guarantees, refactoring any-typed codebases] [Keywords...

Use when designing discriminated unions, debugging control-flow analysis, choosing satisfies vs as, building branded types, writing user-defined type guards, or composing conditional types and template literal types. Triggers: "Type X is not assignable to type Y" after a type guard, exhaustiveness checks via never, satisfies vs explicit annotation, infer in conditional types, mapped types with key remapping, distributive conditional types, type-narrowing inside callbacks losing the narrowed t...

Master typographer specializing in font pairing, typographic hierarchy, OpenType features, variable fonts, and performance-optimized web typography. Use for font selection, type scales, web font optimization, and typographic systems. Activate on "typography", "font pairing", "type scale", "variable fonts", "web fonts", "OpenType", "font loading". NOT for logo design, icon fonts, general CSS styling, or image-based typography.

UI/UX design intelligence. 67 styles, 96 palettes, 57 font pairings, 25 charts, 13 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind, shadcn/ui). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Style...

Friction analysis for anything a person reads or drives - live web flows, wireframes, PDF documents, and LaTeX books - using cognitive psychology, Gestalt grouping, ADHD-friendly design, and a random-surfer reader model that computes where attention pools, where readers backtrack, and where they quit. Covers friction audits, journey and reading simulation, cognitive-load optimization, comprehension-debt auditing (unintroduced ideas, notation working set, entropy cliffs), and Fitts' Law applic...

Coaches behavioral and values-fit interview preparation with negative framing, deep follow-ups, introspection, and mission alignment. Use for culture-fit rounds, Anthropic behavioral prep, failure stories, and self-awareness drilling. Activate on "behavioral interview", "values interview", "culture fit", "tell me about a failure". NOT for coding interviews, system design, resume writing, or technical deep dives.

Apply the foundational taxonomy of workflow control patterns to evaluate, design, and debug complex coordination systems. Based on the seminal "Workflow Patterns" research establishing pattern-based expressiveness evaluation.

Vaporwave + glassomorphic UI designer for photo/memory apps. Masters SwiftUI Material effects, neon pastels, frosted glass blur, retro-futuristic design. Expert in 2025 UI trends (glassmorphism, neubrutalism, Y2K), iOS HIG, dark mode, accessibility, Metal shaders. Activate on 'vaporwave', 'glassmorphism', 'SwiftUI design', 'frosted glass', 'neon aesthetic', 'retro-futuristic', 'Y2K design'. NOT for backend/API (use backend-architect), Windows 3.1 retro (use windows-3-1-web-designer), generic ...

Provision, orchestrate, and cost-fence on-demand GPU rent-a-clusters (Vast.ai, Lambda Labs, RunPod) for heavy mathematical workloads, neural sheaf diffusion, and large-scale Hodge Laplacian spectral solvers. Use when local CPU/GPU is exhausted, when running distributed PyTorch/cuSPARSE batches on complexes with N > 10^5 vertices, or when executing unrolled Sheaf-ADMM parameter optimization. NOT for simple local agent scripting, basic web servers, or non-GPU batch jobs.

Render GPU-accelerated 2D vector graphics (bezier paths, fills, strokes, gradients) and high-quality shaped text in Rust with the Linebender stack: Vello (compute-based vector renderer on wgpu) + Parley (text shaping/layout) + kurbo (geometry) + peniko (brushes/color). Covers winit+wgpu surface setup, building a Scene every frame, feeding Parley glyph runs into Vello, the retained-vs-immediate question, vsync/frame pacing, and the version-pinning minefield between vello/wgpu/parley. Activate ...

Synesthete designer that translates emotional vibes and brand keywords into concrete visual DNA (colors, typography, layouts, interactions). Use when users describe desired "feel" (edgy, trustworthy, premium, playful, minimal) and need specific design specifications. NOT for technical implementation or coding.

Builds a complete, reviewable project plan for a vibe-coded app in July 2026, including user journeys, cold start, account/auth, model/provider readiness, architecture, milestones, tests, launch, and rollback gates. Use when a user asks for a bulletproof plan before or during fast AI-assisted product building. NOT for writing implementation code, generic task lists, or visual design critique without a product plan.

FFmpeg automation for cutting, trimming, concatenating videos. Audio mixing, timeline editing, transitions, effects. Export optimization for YouTube, social media. Subtitle handling, color grading, batch processing. Use for videogen projects, content creation, automated video production. Activate on "video editing", "FFmpeg", "trim video", "concatenate", "transitions", "export optimization". NOT for real-time video editing UI, 3D compositing, or motion graphics.

Use when Vite dev startup is slow, HMR is not invalidating, the build emits chunks larger than expected, or you are authoring a Vite plugin. Triggers: "[plugin:..] failed", "Failed to resolve import", externalize warnings, dependency pre-bundling errors, lib mode + code splitting, ssr externals confusion, "Cannot find package" only in build, manualChunks tuning, optimizeDeps include/exclude tuning, rollup-plugin-visualizer review, --profile flag, configResolved/handleHotUpdate hook authoring....

Mental models and decision frameworks for building autonomous agents that continuously learn, explore, and accumulate skills in open-ended environments without human supervision

Architectural patterns for building agent systems that dynamically decompose complex tasks, generate specialized subagents just-in-time, and prevent cascading failures through adaptive replanning.

Run a declared checkpoint after a whole DAG wave returns, joining node evidence to dependencies, risks, approvals, and a new graph revision before launching a later wave. Use to hold, revise, escalate, or admit a planned next wave when evidence is missing, partial, failed, stale, or changes a predecessor contract. NOT for interrupting a running wave, treating a local RCP-3 sketch as FIPA conformance, or assigning work without authority and resource checks.