hbmartin

Render two brick models under identical fixed views, judge which one looks better against a fixed rubric, record the forced-choice verdict in the project's tracked preference log, and escalate any pair you cannot call confidently to the human. Use when…

Diagnose an LDraw assembly's stability, load paths, connectivity, floating parts, and assembly risks, producing a complete -analysis diagnostics bundle. Use when asked whether a model is stable or buildable, why it might collapse, which bricks are weak or…

Sweep the evaluation corpus (synthetic stress shapes by default, curated meshes opt-in) through the placement strategies, produce an aggregate scorecard in ./legolization-eval/, and diff it against the committed baseline to catch regressions or confirm…

Research, estimate, validate, and activate catalog support for a part the generator does not know yet. Use when a build fails on an unsupported or unknown part, when asked to add or extend part support, infer geometry and mass estimates for an LDraw part…

Audit a model's step-by-step build instructions: machine-check step ordering, per-step stability, insertion pressure, and dangling bricks not yet connected to ground, then read rendered per-step images for sensibility. Use when asked whether instructions make…

Turn a 3D model (.vox, .npy, .obj, .stl, or .ply) into a physically stable brick-built model with a complete portable bundle: winning model, bill of materials, candidate comparison report, and instructions when rendering is available. Use when asked to…

Improve an existing LDraw assembly (.ldr or .mpd): validate and re-publish it exactly as built, or retile it into regenerated brickwork that keeps its shape and colours. Use when asked to optimize, rebuild, retile, strengthen, or reduce the part count of an…

Inspect, orient, scale, colour, and normalize a source model (.vox, .npy, .obj, .stl, or .ply) before brick generation. Use when asked whether an input is usable, which way is up, how big the build will be, what colours were detected, to fix an ambiguous…

Measure and regression-test on-device model output with Apple's Evaluations framework: eval harnesses, prompt hill-climbing, model-as-judge graders and alignment, synthetic or adversarial data, and tool-trajectory scoring; also use DNIKit to audit datasets…

Migrate a shipping Apple AI app or model pipeline from the 26 SDK generation to 27: changed APIs, the Foundation Models Adapter sunset, LanguageModelError taxonomy changes, dual-SDK builds, Core ML-to-Core AI decisions, Metal toolchain requirements, and asset…

Ship and operate on-device AI models in released Apple apps: distribution outside the app binary, background asset packs and updates, download and disk budgets, memory pressure and jetsam, thermal throttling, and honest inference benchmarks. Use when choosing…

Expose app content and actions to Siri and Apple Intelligence with App Intents, assistant schema domains, IntentParameter.valueState, AppEntity, IndexedEntity and Spotlight indexing, on-screen awareness, FileEntityIdentifier and FileRepresentation, display…

Build, convert, optimize, and debug Core AI 27 neural models using AIModel, NDArray, bundles, engines, guided decoding, specialization, caching, AOT compilation, coreai-torch conversion, custom Metal ops, quantization, palettization, pruning, and ANE/GPU…

Build and debug apps using Apple's Foundation Models framework: LanguageModelSession, @Generable, @Guide, streaming, tools, the Instructions-versus-Prompt trust boundary, context and KV cache, DynamicProfile, custom LanguageModel backends, Playground,…

Write and debug hand-built on-device ML kernels with Metal TensorOps and Metal Performance Primitives: MPP and MTLTensor APIs, quantized or multiplane operands, cooperative tensors, threadgroup and memory layout, and flash attention. Use when implementing an…

Build and debug MLX in Python or Swift: mx.array and lazy evaluation, unified memory, mx.compile and transforms, custom kernels, quantization, mlx-lm generation and prompt caching, serving, distributed inference, fine-tuning, model ports, mlx-swift-lm apps,…

Collect Ratchet local pipeline debug evidence. Use when a run is failed, stuck, stale, stopped, missing outputs, has confusing LLM behavior, or needs local evidence from pipeline traces, events.jsonl, worker.log, raw LLM artifacts, and a debug bundle…

Show local Ratchet setup status and optional host-agent configuration.

Evaluate and enhance a ratchet skill using LLM-as-judge A/B testing with an iterative improvement loop.

Run the Ratchet skill extraction pipeline to analyze project sessions and generate reusable skills and strategies.

Show Ratchet help — available commands, output locations, skill and strategy structure, and usage tips.

View or update your Ratchet developer profile (language, level, style) to personalise skill extraction.

Show Ratchet status including pending skills, strategies, and recent pipeline runs.

Stop a running Ratchet pipeline.

Survey a codebase's animation and motion code as a senior motion advisor, then produce a prioritized audit and self-contained implementation plans for other agents (or cheaper models) to execute. Read-only on source code — it plans improvements, it does not…

Reviews animation and motion code against a high craft bar derived from Emil Kowalski's design engineering philosophy. Default to flagging; approval is earned.

Reverse-lookup glossary that turns a vague description of a web animation or motion effect into its exact term ("the bouncy thing when a popover opens" → Pop in; "the iOS rubber-band scroll" → Rubber-banding). Use when the user asks "what's it called when…",…

Apple's approach to interface design and fluid, physical motion, translated for the web. Use when building or reviewing gesture-driven UI, spring animations, drag/swipe/sheet interactions, momentum and interruptible transitions, translucent materials and…

Use when working with iOS/macOS Keychain Services (SecItem queries, kSecClass, OSStatus errors), biometric authentication (LAContext, Face ID, Touch ID), CryptoKit (AES-GCM, ChaChaPoly, ECDSA, ECDH, HPKE, ML-KEM), Secure Enclave, secure credential storage…

Implement, review, or improve SwiftUI animations and transitions. Use when adding explicit animations with withAnimation, configuring implicit animations with .animation(_:body:) or .animation(_:value:), configuring spring animations (.smooth, .snappy,…

Inspect SonarQube Cloud/SonarCloud findings for this repository using local .env credentials. Use when the user asks to inspect, summarize, triage, prioritize, or fix SonarQube/SonarCloud issues, quality gate results, security hotspots, vulnerabilities, bugs,…

Fetch GitHub pull request feedback for a PR number and triage each comment/review item as already fixed, high-level, should be fixed, or should not be fixed. Use when the user asks to evaluate GitHub PR feedback status from `gh pr view`.

Emit clear progress updates by default for most non-trivial tasks and speak milestone status. Use whenever work has multiple phases, may include long-running commands or tests, involves waiting on tools or agents, or benefits from hands-free audible status.

Transform any photo into surrealist AI art. Uses Claude to write a story about your photo, then Gemini generates a reimagined version in one of 35+ visual styles. Deep Apple Photos integration — pick from albums, save back, schedule as an Apple TV screensaver.

Turn your Overcast listening history into actionable intelligence. Syncs episodes, transcripts, and chapters to SQLite, then uses LLM analysis to surface insights from what you've listened to and connect them to your current projects and interests. Depth of…

Send a notification after work completes. Use when the agent should alert you about task completion or important milestones.

Inspect SonarQube Cloud/SonarCloud findings for this repository using local .env credentials. Use when the user asks to inspect, summarize, triage, prioritize, or fix SonarQube/SonarCloud issues, quality gate results, security hotspots, vulnerabilities, bugs,…

Visually verify graphviz2drawio conversion fidelity by rendering a graphviz source side-by-side as a native graphviz PNG and a converted draw.io PNG, then reading both images to compare. Use this whenever modifying conversion code in graphviz2drawio/ (nodes,…

Visually verify graphviz2drawio conversion fidelity by rendering a Graphviz source as both native Graphviz PNG and converted draw.io PNG, then inspecting the images. Use when modifying graphviz2drawio conversion or rendering code, updating specs, fixing…

Prepare, validate, upload, and verify CREG's internal-only TestFlight Beta build. Use when asked to preflight, archive, export, publish, upload, or confirm the processing status of a CREG TestFlight build.

Use SwiftUI #Preview as the default visual harness when implementing or reviewing UI from screenshots, mockups, Figma frames, or design specifications. Create deterministic named preview states, render and compare them, and avoid launching the full app in…

Primary W&B skill for broad or mixed Weights & Biases work: project overviews, W&B runs and artifacts, Weave traces and evaluations, Reports, and Launch workflows. Use when the task spans multiple W&B surfaces or the user asks generally what is happening in a…

Safely convert or incrementally extract first-party gitx production Objective-C into Swift while preserving observable behavior and Objective-C interoperability. Use whenever Codex explicitly converts Objective-C, assesses a conversion candidate, or…

Plan, add, run, and review XCTest-based verification for GitX. Use whenever changing GitX behavior, fixing a bug, adding or reviewing tests, diagnosing test, coverage, analyzer, sanitizer, or performance failures, changing shared test plans or CI…

Search Fireflies.ai meeting transcripts by keyword and download the full matching transcripts as Markdown. Use when the user wants to find, look up, pull, or read meeting notes or call transcripts from Fireflies by topic, phrase, speaker, or keyword (e.g.…

Inspect SonarQube Cloud/SonarCloud findings for this repository using local .env credentials. Use when the user asks to inspect, summarize, triage, prioritize, or fix SonarQube/SonarCloud issues, quality gate results, security hotspots, vulnerabilities, bugs,…

Inspect SonarQube Cloud/SonarCloud findings for this repository using local .env credentials. Use when the user asks to inspect, summarize, triage, prioritize, or fix SonarQube/SonarCloud issues, quality gate results, security hotspots, vulnerabilities, bugs,…

Inspect SonarQube Cloud/SonarCloud findings for this repository using local .env credentials. Use when the user asks to inspect, summarize, triage, prioritize, or fix SonarQube/SonarCloud issues, quality gate results, security hotspots, vulnerabilities, bugs,…