- Skills
- 29
- 仓库
- 1
- 职业覆盖
- 0
- 场景覆盖
- 0
- Stars
- ★ 1
Sekou Doumbouya 的 Skills 29
voice
暂无简介
workbench
暂无简介
workbench
暂无简介
workbench
暂无简介
anvil-serving-voice-ops
Operate topology-owned anvil-serving voice audio and Realtime proxy lifecycle, Mini-to-Dark loopback forwarding, profiles, sidecars, and bounded voice benchmarks. Use for voice status, logs, start/stop/restart, OpenClaw Talk topology checks, and voice evidence that must remain separate from LLM serve qualification evidence.
anvil-serving-stt-benchmark
Prepare, validate, execute, compare, and publish reproducible multi-sample speech-to-text qualifications in anvil-serving. Use for STT model bakeoffs, corpus-based WER/CER and latency testing, candidate overlays, concurrency tests, restoration checks, and fail-closed stt-benchmark-evidence/v1 artifacts. Do not use for a single voice-pipeline turn or LLM qualification.
anvil-serving-release-readiness
Gate an Anvil Serving package release and coordinated live deployment across Windows, Primary Node, Companion Node, the controller, router, Hermes, Pi, and OpenClaw. Use when a change will be merged, tagged, published, or deployed and exact manifest dependencies, version parity, recovery policy, and outage-free client smokes must be proven.
anvil-serving-recipe-feasibility
Bound and eliminate local-model serve recipes with reproducible interval math before expensive downloads or benchmarks. Use for VRAM/RAM/context feasibility, KV-cache and speculative-draft budgeting, quant/runtime shortlists, throughput or time-to-success thresholds, intelligence-loss budgets, and campaigns where unknown serving variables must remain explicit and be updated from evidence.
anvil-serving-llm-qualification
Qualify a pinned local LLM or VLM on Anvil Serving with reproducible text, image, video, capacity, router, and storage evidence. Use for model bakeoffs, failure diagnosis and researched configuration trials, quant comparisons, context/concurrency qualification, multimodal enablement, or production-role recommendations that must remain human-gated.
anvil-serving-kernel-tuning
Decide whether an Anvil Serving model needs hardware-specific kernel tuning, run an isolated engine tuner, qualify the result against an untuned baseline, and recommend, store, or reject the tune with reproducible evidence. Use for missing MoE/GEMM kernel-config warnings, suspected kernel bottlenecks, engine or GPU upgrades that invalidate an existing tune, and default-versus-tuned throughput or latency comparisons.
anvil-serving-benchmark-docs
Publish and maintain hardware-first, publication-ready benchmark documentation in anvil-serving. Use whenever a model is newly configured, served, benchmarked, qualified, promoted, rolled back, or fails meaningfully, including LLM, vision, Omni, STT, and TTS results. Separates measured hardware from protected/co-resident topology and local evidence from research priors; never authorizes a serve or route change.
anvil-observatory-validation
Validate an installed Anvil Observatory console, its authentication boundary, packaged assets, and read-only data coverage with a bounded evidence receipt.
anvil-media
Generate, inspect, or cancel images and videos through Anvil's bounded named-workflow MCP tools.
anvil-serving-llm-qualification
Qualify a pinned local LLM or VLM on Anvil Serving with reproducible text, image, video, capacity, router, and storage evidence. Use for model bakeoffs, failure diagnosis and researched configuration trials, quant comparisons, context/concurrency qualification, multimodal enablement, or production-role recommendations that must remain human-gated.
anvil-serving-benchmark-docs
Publish and maintain hardware-first, publication-ready benchmark documentation in anvil-serving. Use whenever a model is newly configured, served, benchmarked, qualified, promoted, rolled back, or fails meaningfully, including LLM, vision, Omni, STT, and TTS results. Separates measured hardware from protected/co-resident topology and local evidence from research priors; never authorizes a serve or route change.
anvil-media
Generate, inspect, or cancel images and videos through Anvil's bounded named-workflow MCP tools.
anvil-serving-workbench
Operate anvil-serving from an agent harness using MCP/controller-backed playbooks, CLI fallbacks, and sub-agent evidence packets. Use when asked to inspect Anvil Serving status, choose models, run preflight or benchmarks, sync OpenClaw/Codex/Claude Code workbench config, collect promotion evidence, or coordinate small-model sub-agents for anvil-serving operations.
anvil-serving-voice-ops
Operate topology-owned Anvil Voice lifecycle, Mini-to-Dark audio routing, sidecars, profiles, proxies, and bounded voice benchmarks. Use for voice status, logs, start/stop/restart, OpenClaw Talk checks, and voice evidence.