toxicwind
@toxicwind
Public Skills
omni-to-hodos-cutover
by toxicwind
Procedure for cutting over tau agent configuration from obsolete OmniRoute to VansRouter/hodos gateway system
tau-extension-location
by toxicwind
Remember where tau extensions like tau-hodossync actually live vs where their configs are stored
tau-extension-config-rules
by toxicwind
How to register tau user extensions and resolve tau config layering issues
nemotron-nano3
by toxicwind
Reference desk for Nemotron 3 Nano / Llama-Nemotron Nano 3 — architecture, training data, recipes, evaluation, quantization, deployment. Use when the architect asks facts about the model rather than building a pipeline.
nemotron-retrieval-recipes
by toxicwind
Use when planning, debugging, tuning, evaluating, exporting, or deploying public Nemotron embed/rerank retrieval recipes.
nemotron-add-step
by toxicwind
Add a new step under src/nemotron/steps/<category>/<step_id>/ — manifest (step.toml), runner glue, configs, and per-step README.md. Use when extending the catalog so /nemotron-customize can route to it.
smithers
by toxicwind
Drive Smithers, a durable control plane for long-running coding agents. Use when the architect wants multi-step, long-running, crash-safe, or human-in-the-loop agent work: "orchestrate agents", "run a workflow", "implement this and review it", "keep iterating until tests pass", "plan then build", or anything that needs retries, approvals, replay, or evals across multiple AI steps. YOU (the agent) run Smithers on the user's behalf; it is not a GUI the human clicks. HARD RULE 0: if SMITHERS_INSIDE_RUN is set in your environment you are ALREADY a worker agent inside a Smithers node, so never invoke the Smithers CLI or MCP tools to launch or steer a run; do the node's task directly with your ordinary tools. Everything below applies only when it is unset. You are otherwise an ORCHESTRATOR: run long-running, multi-step, or background work through Smithers, not through your own ad-hoc subagents; spend your time observing the run and reporting. Also governs DELIVERABLE SHAPE even when no workflow is involved: an architect-facing report, plan, or architecture document is a self-contained HTML page, and workflow UIs are composed from the shipped shared components (MarkdownEditor, DiffHunks, chat surfaces, KpiStat, StatusPill, EmptyState) rather than hand-rolled markup.
nemotron-customize
by toxicwind
"Plan, configure, and chain repo-native Nemotron customization steps into single-step or multi-step pipelines: curation, translation, SFT/PEFT (AutoModel or Megatron-Bridge), pretraining/CPT, RL alignment (DPO/RLVR/GRPO/RLHF), BYOB/MCQ benchmarks, checkpoint conversion, ModelOpt optimization, env profiles, and evaluation of trained checkpoints or existing/hosted endpoints. Use when a request names a Nemotron step or workflow, or asks to clean, translate, train, fine-tune, align, convert, optimize, evaluate, or compose these into a pipeline. Do NOT use for frontend/dashboard/visualization work, generic ML advice, billing/access, or non-Nemotron coding tasks."
nemotron-ultra
by toxicwind
Reference desk for NVIDIA Nemotron 3 Ultra (550B-A55B) — architecture, NVFP4 pretraining, SFT, MOPD (multi-teacher on-policy distillation), MTP boosting, quantization, inference. Use when the architect asks facts about Ultra rather than building a pipeline.
nemotron-super3
by toxicwind
Reference desk for NVIDIA Nemotron 3 Super — architecture, training data, recipes (pretrain/SFT/RL/eval/quantization), and deployment notes. Use when the architect asks facts about Super3 rather than building a pipeline.
airspeak
by toxicwind
Write and review technical documentation — docs, READMEs, PR descriptions, error messages, runbooks, release notes, changelogs. English is inspired by ASD-STE100 Issue 9. Use when the architect writes or edits technical prose, API docs, or any document where clarity and brevity matter — even if they don't explicitly mention a style guide or simplification. Never use for marketing copy, essays, creative writing, or anything that needs a distinctive voice. These specs intentionally strip voice.
nemotron-policy-generator
by toxicwind
"Generates BYO custom safety policies for NVIDIA Nemotron content-safety guardrails — Nemotron-Content-Safety-Reasoning-4B (text) and multimodal Nemotron-3-Content-Safety. Produces a Markdown policy, JSON taxonomy, and drop-in inference prompts. Maps rough words or an existing policy to V2 categories, adding custom categories or topic-following rules."
tau-fork-pinning
by toxicwind
"Procedure for syncing, building, and immutably pinning Tau fork binaries."
ast-bm25-racer
by toxicwind
Estate AST-BM25 Racer, local-heavy edition: hybrid code retrieval plus nanosecond candidate racing. Triggers on: code retrieval, candidate racing, AST search.
phone-backup
by toxicwind
Bulk backup and storage triage for User's Pixel 9 Pro XL via ADB on yote. Benchmarked fastest pull (parallel adb pull, NOT tar-stream), phone storage map, the toybox find quirk, 8TB archive layout, and the future-forward Syncthing continuous-sync direction. Use when User asks about phone storage, pulling files off the phone, or freeing phone space. Photos (DCIM/Pictures) are excluded by standing direction — this skill handles code, archives, repos, and documents only.
nemotron-add-pattern
by toxicwind
Add a cross-cutting decision pattern under src/nemotron/steps/patterns/. Use when a recurring ML decision (tokenizer lock, eval bookends, LoRA-on-small-data, etc.) must be encoded so other skills can fire it during planning.
nemotron-add-model
by toxicwind
Onboard a new model family (Nemotron or third-party) into skills/ — paper chunks, recipe summaries, context packs, and model card. Use when a contributor wants downstream skills like /nemotron-customize to be able to route to a new model.
fix-gatehouse-mcp-auth
by toxicwind
Fix MCP gatehouse authentication by setting require_mcp_auth to false in config file
estate-submodule-update
by toxicwind
Procedure for adding a submodule to estate repository and updating README with infrastructure changes
classifier-project
by toxicwind
Internal meta tool for classifier false-positive repair. Relaunch of classifier-safe-docs as a first-class project: audit, rewrite, and verify docs and task bodies so safety classifiers stop flagging benign estate operations. Behavioral language only. Use when any doc, task body, or agent brief gets flagged, when lanes stall on classifier refusals, or when Chris says "make it not trigger the classifier".
exa
by toxicwind
"Use Exa when the user asks for Exa, neural web search, or this provider's API."
paper-search
by toxicwind
Paper research as a first-class chat capability: PAPER-TASK/PAPER-RESULT protocol over the fleet channel, arXiv + alphaXiv leg racer, pitchfork-managed poller daemon. Triggers on: "paper search", "arxiv", "paper-poller".