SOP for debugging browser automation failures on complex websites. Use when browser tools fail on specific sites like LinkedIn, Twitter/X, SPAs, or sites with Shadow DOM.

Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.

Reduce LLM size and accelerate inference using pruning techniques like Wanda and SparseGPT. Use when compressing models without retraining, achieving 50% sparsity with minimal accuracy loss, or enabling faster inference on hardware accelerators. Covers unstructured pruning, structured pruning, N:M sparsity, magnitude pruning, and one-shot methods.

OpenAI's model connecting vision and language. Enables zero-shot image classification, image-text matching, and cross-modal retrieval. Trained on 400M image-text pairs. Use for image search, content moderation, or vision-language tasks without fine-tuning. Best for general-purpose image understanding.

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

Stage and create a Conventional Commits message for the current change

Multi-page Lighthouse audit via the MIT-licensed Unlighthouse CLI. Free-tier alternative to running PageSpeed against every URL on a site, no API quota burn, runs locally.

Use only when a focused Sales workflow has selected a present and connected ZoomInfo connector, or the user explicitly asks for ZoomInfo company search, contact search, enrichment, intent, similarity, recommendations, or research. Do not use when another enrichment source is selected.

Use when the user wants a competitor or vendor comparison, market-landscape analysis, battlecard, objection package, positioning brief, or account-specific competitive view. Produce an evidence-backed comparison, guidance, and brief, using supplied materials, connected research, and public evidence when appropriate.

Build a source-backed dashboard or monitoring view with clear metrics, filters, visual hierarchy, definitions, refresh context, and QA.

Understand and customize the local Trellis architecture inside a user project. Use when modifying .trellis plus platform hooks, settings, agents, skills, commands, prompts, workflows, the channel runtime (trellis channel), bundled runtime agents under .trellis/agents/, selectable workflow templates, registry-backed spec refresh, cross-session memory (trellis mem) generated by trellis init, or AI-facing bundled skills (trellis-channel, trellis-session-insight, trellis-spec-bootstrap) and bundled-skill auto-dispatch flow.

Formulate statistical research problems with formal notation, target parameters, assumptions, hypotheses, evaluation criteria, and theory targets.

Design and run statistical experiments that test the formal problem, proposed methods, theoretical predictions, baselines, and ablations.

Run Flux Balance Analysis (FBA) and related constraint-based simulations using COBRApy. Covers standard FBA, parsimonious FBA (pFBA), Flux Variability Analysis (FVA), loopless FBA, gene/reaction knockouts, and carbon source swapping. Outputs flux distributions and CSV files.

Runs batch and HPC workloads on GKE, utilizing job queues and parallel processing. Use when running GKE batch jobs, configuring GKE HPC, or setting up GKE job queues. Don't use for standard web application deployments (use gke-app-onboarding instead).

Manages Agent Platform serving endpoints. Use when you need to create, list, describe, update, or delete serving endpoints for model deployment on Agent Platform. Also use when troubleshooting endpoint permission, quota, or resource busy errors. Don't use for deploying models to endpoints or for running model evaluations.

Slate v2 Autoresearch recipe picker. Lists/recommends Codex Autoresearch recipes, produces read-only setup plans, and maps non-perf loops such as test runtime, typecheck, bundle size, memory, command latency, and quality-gap.

Resolve one Slate or Slate v2 GitHub issue end to end, using task-style issue intake plus Slate Plan proof discipline, then post the verified no-PR issue comment with gh.

Pi-compatible PUA high-agency governance skill. Use for explicit PUA/PIP requests, repeated failures, passive/giving-up behavior, user frustration, or unverified completion. Pair with /pua-on extension command when persistent pressure is desired.

Find user-visible changes merged to NemoClaw and update their owning documentation. Use in the post-merge documentation workflow or for direct documentation catch-up. Derive pages, commands, variants, and validation from the current checkout. Trigger keywords - update docs, docs from commits, catch up docs, docs drift.