Profile My skills Saved Collections Edit profile Submit a skill

← all publishers

steveseguin

@steveseguin source repo

7 published skills

  1. ▌
    Ninja P2p 2 · steveseguin
    Use the installed ninja-p2p CLI when Claude should coordinate with room peers, keep or wake a sidecar inbox, exchange files, run resumable swarm transfers, or bridge Social Stream Ninja over WebRTC.
    0
    installs
  2. ▌
    MCP · steveseguin bundle
    Use VDO.Ninja WebRTC data channels as MCP tools for bot-to-bot messaging and reliable file transfer, with clear transport limits, optional TURN, and optional join-token/session-MAC security.
    0
    installs
  3. ▌
    Ninja P2p · steveseguin bundle
    Use the installed ninja-p2p CLI when the user wants Codex to coordinate with room peers, keep or wake a sidecar inbox, exchange files, run resumable swarm transfers, or bridge Social Stream Ninja over WebRTC.
    0
    installs
  4. ▌
    Add Model · steveseguin
    Add or update an LLM preset in MODEL_PRESETS (engine.js) from its official config.json so the planner's physics (KV cache, FLOPs, fixed per-layer overhead) are right for it; wires the picker, the evidence rules, and the tests. Use whenever a user asks to add/support/update a model (Qwen, Gemma, DeepSeek, Llama, Mistral, Kimi, GLM, MiniMax, Nemotron, Muse, Granite, LFM…) or to refresh a preset's specs.
    0
    installs
  5. ▌
    Add Hardware · steveseguin
    Add or update a device template in DEVICE_TEMPLATES (engine.js) — GPU, Apple/AMD unified-memory system, CPU, NVMe tier, or accelerator — with official peak specs, the right backend overhead scale, default runtime, picker group, and evidence mapping. Use when a user asks to add/support/fix a GPU, Mac, DGX, Strix Halo, Arc, Instinct, TPU, CPU, or storage tier.
    0
    installs
  6. ▌
    Calibrate Engine · steveseguin
    Re-fit or re-anchor the planner's decode/prefill physics (FRAMEWORK_PROFILES, LAYER_OVERHEAD_SCALES, device kernelOverheadScale) against the gold-case corpus, triage outliers and roofline violations, and re-pin the exact-value regression tests. Use when predictions look wrong for a model/runtime/hardware class, when `npm run audit:gold` fails, or after changing any engine formula.
    0
    installs
  7. ▌
    Refresh Evidence · steveseguin
    Refresh the community benchmark snapshot (data/localmaxxing-snapshot.js) from the public Localmaxxing API, understand how gold cases are selected and what the rows mean, triage a failing weekly refresh, and map new models/hardware so their runs become evidence. Use when the snapshot is stale, the CI refresh workflow fails, a new model/device has community runs, or a gold row looks wrong.
    0
    installs