176 skills · sorted by relevance
Guides adapting OpenPI robot models to custom datasets and serving them to produce actions from observations. Covers JAX and PyTorch training, checkpoint conversion, and troubleshooting for normalization statistics and GPU memory.
Guides PyTorch model training and fine-tuning, from choosing architectures to configuring optimizers and learning rates. Provides troubleshooting steps for unstable loss and memory shortages, plus methods for improving GPU throughput and comparing experiments.
Add PyTorch FSDP2 to training scripts so model state can be split across GPUs. Covers distributed initialization, mixed precision, CPU offloading, and checkpoint saving and loading for models that exceed a single GPU's memory.
Write and debug code for further training large language models with TRL. Match training methods to dataset formats, configure supervised or preference-based training, and set up reward functions through Python or the command line.
Pick an LM output format per (task x consumer x model) rather than by reflex: different formats carry different cognitive load (e.g. code-in-JSON makes the same model write worse code than plain-text+diff, while asking for prose when you need a typed object fails the other way). Use when designing or debugging an LM's output schema, choosing between plain text / diff / JSON / tool-call / grammar-constrained output, or when a model's quality drops after wrapping its output in a structured format.
The English summary is being prepared. Showing the original description.
Main app ViewModel patterns (ScopedViewModel, Hilt, StateFlow, triggerEvent, navArgs, savedState flows, events). Use when writing, editing, exploring, debugging, or reviewing ViewModels in the store management app. Covers navArgs, combine+asLiveData, sealed state hierarchies, custom events, init blocks, and companion object placement. NOT for POS (WooPos*) code — use the `pos` skill instead.
The English summary is being prepared. Showing the original description.
Use when working with Bentoml — bentoML model serving and packaging management. Covers service management, model packaging, deployment status, API testing, Bento building, and runner configuration. Use when packaging ML models for serving, deploying BentoML services, debugging inference issues, or managing model artifacts.
The English summary is being prepared. Showing the original description.
Expert guidance for designing, implementing, migrating, and debugging SwiftData persistence in Swift and SwiftUI apps. Use when working with @Model schemas, @Relationship/@Attribute rules, Query or FetchDescriptor data access, ModelContainer/ModelContext configuration, CloudKit sync, SchemaMigrationPlan/history APIs, ModelActor concurrency isolation, or Core Data to SwiftData adoption/coexistence.
The English summary is being prepared. Showing the original description.
Use when designing, running, or judging an experiment on a Syntropic137 workflow - comparing prompt variants, fan-out vs single agent, adversarial cross-model review, model selection (can Haiku do this job), phase splits, or tool grants. Trigger phrases include "run a workflow experiment", "A/B the workflow", "does fan-out help", "which model for this phase", "is this prompt better", "measure review quality", "why did the workflow do that". Do NOT use for a one-off workflow run that is just doing work, for authoring a workflow that is not being compared against anything, or for debugging a single failed execution (that is ordinary debugging).
The English summary is being prepared. Showing the original description.
Diagnose OpenClaw problems involving models, providers, tools, and streamed responses. Use targeted logs, repeat the failing operation, and inspect stored conversations and attachments to identify the cause, guide a repair, and verify the result.
Helps you use Unsloth for language model fine-tuning with LoRA and QLoRA, methods that adapt models with fewer trainable parameters. Supports API questions, implementation, and debugging using reference material drawn from official documentation.
Configure supervised fine-tuning for models that process images and text. Define which components remain unchanged, set LoRA adapters for updating selected model weights, choose image resolution limits, and investigate training runs whose evaluation results fail to improve.
Get guidance on adapting LLMs with Unsloth, including implementation, API usage, and debugging. Covers LoRA and QLoRA, methods for adjusting models with fewer trainable parameters, with a focus on training speed and memory use.
Get guidance on adapting large language models with Axolotl. Covers YAML configuration files, LoRA and QLoRA methods for training fewer parameters, distributed training settings, dataset formats, and debugging training code.
Investigate difficult bugs by adding temporary logs to the actual execution path and reproducing the reported behavior. Analyze JSONL logs, which store one structured event per line, then fix the cause, verify regression tests, and remove debugging artifacts.
Write integration tests for core error processors by simulating API failures with a mock model. Check retry attempts, changes to conversation messages, and streamed responses, with guidance on test setup and diagnosing empty output.
Investigate difficult bugs through explicit hypotheses, targeted logs, and repeated checks against actual data. Keep an investigation journal to track findings, examine environment differences and leftover state, and confirm the root cause before applying a fix.
Guides fine-tuning and evaluation of OpenVLA-OFT robot models that turn images and language into actions. Covers LIBERO simulation, ALOHA deployment, LoRA adaptation that updates a small set of parameters, and troubleshooting action scaling or configuration mismatches.
Test Omnigent’s GitHub Copilot SDK integration on a live local server. Build test agents, run real conversations, and inspect authentication, model selection, tool connections, policy enforcement, and leftover processes to investigate integration failures.
Test Omnigent’s native Pi integration using the real terminal interface and a local server. Verify messages sent through the web or API, recorded replies, interruption, session resumption, authentication, and model settings to investigate integration problems.
Test Omnigent’s Antigravity SDK integration against a live local server. Build test agents and run real conversations to check Gemini authentication, model selection, tool connections, and failures across the full execution path.
Test Omnigent’s Cursor SDK integration against a live local server. Build test agents, run real conversations, and check authentication, model selection, and tool connections while investigating empty responses or bridge processes left after shutdown.
Review how an AI agent loads, retains, summarizes, and shares working information. This historical guide helps examine forgotten constraints, repeated actions, tool selection, and handoffs using context engineering practices published in 2024–2025.
Visualize training metrics, debug models with histograms, compare experiments, visualize model graphs, and profile performance with TensorBoard - Google's ML visualization toolkit
The English summary is being prepared. Showing the original description.