Writes and debugs TorchDrug 0.2.1 code for molecular property prediction, generation, retrosynthesis, protein and knowledge-graph learning.
A skill for building, differentiating and training quantum circuits with PennyLane across IBM, Google, Rigetti and IonQ backends.
Version-aware guidance for PufferLib: wrapping environments, vectorization, PuffeRL training plans, and safe checkpoint review across the 3.0.0 release and the 4.0 source line.
Restructures a bloated CLAUDE.md / AGENTS.md into a lean Level 1 plus on-demand Level 2 references, with verified zero information loss.
An evidence-driven, falsification-first playbook for network, streaming and protocol bugs where the obvious cause is probably wrong.
Measure any OpenAI- or Anthropic-compatible endpoint across six dimensions — availability, request fidelity, speed, concurrency, protocol compliance, and quality regression.
A battle-tested guide to writing, testing, registering and debugging Claude Code PreToolUse/PostToolUse/SessionStart/Stop hooks.
A meta-skill for authoring new Claude Code skills, upgrading existing ones, and proving the change worked with tiered evals.
A field-tested guide for building iOS/macOS apps with XcodeGen, SwiftUI and SPM, plus Apple signing, notarization and CI/CD fixes.
Turns any Claude Code skills repo into an official plugin marketplace that users can install with `claude plugin install`.
Audits already-rendered web, dashboard, HTML deck, and Electron UIs with real-browser journeys, inspected screenshots, and a bundled Playwright sweep to catch layout, responsive, state, and output defects.
Spins up 5 mutually hostile agents for a 3-round critique debate, keeps only the defensible insights, then hands them to the `plan` agent for an executable plan.
Maps your codebase into per-feature Mermaid flowcharts, proves where logic is duplicated, and proposes the simplest unified architecture.
Type 'ultrawork' in your prompt and the agent switches to a maximum-precision mode that demands failing-first proofs, real-surface evidence, and cleanup receipts before claiming done.
Analyzes your whole repo and generates a hierarchical AGENTS.md knowledge base at the root plus high-complexity subdirectories.
A QA-focused skill for smoke-testing opencode's CLI, HTTP server, plugin hooks/events and TUI, plus read-only session DB investigation.
Typing 'ultrawork' or 'ulw' flips the agent into a maximum-precision mode that demands failing-first proofs, real-surface QA evidence, and cleanup receipts before it can claim done.
Turns a vague or large request into one decision-complete work plan: it explores the codebase first, asks only the real forks, and writes the plan only after your explicit okay.
A six-phase refactoring workflow that forces codemapping, coverage checks, and test verification after every single change.
Turns a vague or large request into one decision-complete work plan: explores the codebase first, asks only the real forks, and writes the plan only after your explicit okay.