[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77RJSYENH9ERF4RB3P8Y0DY":3,"skill-same:01M77RJSYENH9ERF4RB3P8Y0DY":46},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M77RJSYENH9ERF4RB3P8Y0DY","anvil-serving-voice-ops",null,"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\u002Fstop\u002Frestart, OpenClaw Talk topology checks, and voice evidence that must remain separate from LLM serve qualification evidence.","fakoli\u002Fanvil-serving","fakoli","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1631326?v=4","01M77RFVS9JSVF3PFEJ4114QDT","Sekou Doumbouya","01M77RFVT2YT7DTK045N6XFE5G",1,"MIT","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Fskills\u002Fanvil-serving-voice-ops\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: anvil-serving-voice-ops\ndescription: 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\u002Fstop\u002Frestart, OpenClaw Talk topology checks, and voice evidence that must remain separate from LLM serve qualification evidence.\n---\n\n# Anvil Serving Voice Ops\n\nUse this skill for voice-pipeline operations: validating sidecar manifests,\nrendering speech-to-speech commands, managing topology-owned STT\u002FTTS, managing the\npersistent Mini Realtime proxy, selecting profiles, running the Mini-local\nforwarding bridge, and collecting bounded voice benchmarks.\n\n## Start Here\n\n1. Read `README.md`, `CLAUDE.md`, and\n   `docs\u002FOPERATOR-SKILLS-AND-SUBAGENTS.md` before changing behavior or\n   reporting voice results as product evidence.\n2. Prefer `voice_manage` for structured STT\u002FTTS lifecycle and\n   `voice_proxy_manage` for persistent Realtime proxy lifecycle. Use the\n   existing CLI verbs for foreground, profile, sidecar, bridge, and benchmark\n   operations that do not have MCP wrappers:\n   `anvil-serving voice sidecar validate`,\n   `anvil-serving voice sidecar command`,\n   `anvil-serving voice sidecar compose`,\n   `anvil-serving voice audio {up|down|status|logs}`,\n   `anvil-serving voice profiles {list|validate}`,\n   `anvil-serving voice proxy {run|up|down|restart|status|logs|bridge}`, and\n   `anvil-serving voice benchmark`.\n3. Use `127.0.0.1` for local URLs. Do not introduce `localhost`.\n4. Pass secrets by environment variable name only. Do not put literal API keys\n   in manifests, commands, compose snippets, packets, or logs.\n5. Record the command host before interpreting loopback. In the reference\n   OpenClaw topology, Companion Node runs Gateway plus Anvil Voice Realtime\u002Fproxy\n   and reserves its 16 GB RAM for OpenClaw, Claude Code, and Codex. Do not run\n   STT\u002FTTS\u002FLLM model serves on Mini for reference testing. Use `dark-audio` for\n   Dark bridge ports or `mini-dark-audio-proxy` for Mini proxy ports forwarding\n   to Dark. `mini-audio` is an explicit optional same-host\u002Flocal-audio mode\n   only. A non-gateway checkout cannot validate Mini-local proxy paths by\n   calling its own `127.0.0.1`.\n6. Treat voice benchmark output as voice-pipeline evidence. Voice results do\n   not qualify an LLM serve or change a capability alias.\n7. Route multi-sample STT corpus qualification to\n   `skills\u002Fanvil-serving-stt-benchmark\u002FSKILL.md`. Keep this skill for\n   single-turn audio\u002Fvoice-pipeline evidence and topology lifecycle.\n8. Do not introduce one-off lifecycle, proxy, or port-forwarding scripts as the\n   operational path. If repeatable voice operation needs a new control, it\n   belongs in `anvil-serving voice` and, when agent-operated, MCP\u002Fcontroller.\n\n## Gates\n\nStop for a human gate before public or non-loopback Realtime binds, exposing a\nRealtime token, changing\ncloud\u002FSTT\u002FTTS spend, installing or updating external speech-to-speech packages,\nor starting long-running foreground services on a shared host.\n\n`voice sidecar command --with-auth` and `voice sidecar compose --with-auth`\nrender environment variable references for authenticated router calls. The\ncommand path returns argv, so explicitly warn that the env-var reference appears\nin process argv. The compose path includes an auth exposure comment in the\nrendered service. Use either path only on private hosts where process and\nDocker metadata are protected.\n\nNever use `anvil-serving voice benchmark` as a shortcut for LLM serve\nqualification. Voice benchmarks measure the voice pipeline: turn latency,\nTTFA, audio\u002FSTT\u002FTTS behavior, realtime transport, and user-perceived loop\nquality. They can support a voice-pipeline status report, but an LLM serve\nstill requires its own preflight, benchmark evidence, independent review, and\nhuman approval before an operator changes a capability alias.\n\nUse `workflow_packet_validate` before handing a voice packet to the broader\nworkbench. Voice artifacts must remain explicitly marked as non-promotion\nevidence.\n\n## Playbooks\n\n- Sidecar manifest validation:\n  `anvil-serving voice sidecar validate --config \u003Cmanifest>`.\n- Host command rendering:\n  `anvil-serving voice sidecar command --config \u003Cmanifest> --json`; add\n  `--with-auth` only when the operator accepts the argv exposure warning.\n- Compose rendering:\n  `anvil-serving voice sidecar compose --config \u003Cmanifest>`; inspect loopback\n  ports, `host.docker.internal` routing, image name, and auth comments.\n- Managed audio serves:\n  prefer `voice_manage` for agent\u002Fcontroller operation. The CLI equivalents are\n  `anvil-serving voice audio up --topology \u003Ctopology> --config \u003Cmanifest>` and\n  `voice audio down`; preview mutations first and require `--confirm` for live\n  execution. Use `anvil-serving voice audio status` and bounded\n  `anvil-serving voice audio logs` for reads.\n  External lifecycle serves\n  should be reported as skipped rather than forced into local Docker control.\n- Profile selection:\n  `anvil-serving voice profiles list --config \u003Cmanifest>` to list profiles,\n  `anvil-serving voice profiles validate --config \u003Cmanifest> --profile \u003Cname>`\n  before live work, then add `--profile \u003Cname>` to audio, proxy, or\n  `voice benchmark` when switching audio topology. For reference OpenClaw Talk\n  and candidate benchmarks, keep Mini model-free and select Dark-host audio or\n  a Mini-side proxy to Dark. Use `mini-audio` only when explicitly testing the\n  optional same-host\u002Flocal-audio mode.\n- Mini-to-Dark audio bridge:\n  `anvil-serving voice proxy bridge --topology \u003Ctopology> --config \u003Cmanifest>\n  --profile mini-dark-audio-proxy --dry-run` first. Run it on Mini. Require\n  topology-derived Dark targets and reject every non-loopback Mini listener.\n- Realtime proxy:\n  prefer `voice_proxy_manage` for the persistent Mini process. The CLI\n  equivalents are `anvil-serving voice proxy up` for persistent operation and\n  `anvil-serving voice proxy run` for a foreground diagnostic. Use\n  `down`\u002F`restart` only after preview and explicit\n  confirmation; use bounded `status`\u002F`logs` for inspection. Treat unreachable\n  STT\u002FTTS\u002Frouter endpoints as blockers. Never start audio lifecycle implicitly.\n- Bounded benchmark:\n  `anvil-serving voice benchmark --config \u003Cmanifest>` and capture JSON as\n  `voice-pipeline` evidence with `promoted=false`. The benchmark is an endpoint\n  client, not a Realtime-proxy lifecycle operation, so it has no proxy owner.\n  Use `--scope audio` for an LLM-free TTS-to-STT loop. Keep audio\n  topology and LLM selection separate: `--profile` selects the audio path and\n  `--candidate-overlay \u003Ctoml>` selects a candidate LLM configuration. Do not\n  combine `--candidate-overlay` with direct candidate flags. A voice benchmark\n  does not replace `anvil-serving eval preflight`, capacity benchmarking, or\n  repeated `anvil-serving eval benchmark quality` evidence for LLM serve\n  qualification.\n  Publish a meaningful retained result through\n  `skills\u002Fanvil-serving-benchmark-docs\u002FSKILL.md`; keep measured hardware\n  separate from protected\u002Fco-resident hardware and retain `promoted=false`.\n  Apply its `references\u002Fartifact-set-contract.md` and\n  `anvil-serving.benchmark-artifact-set\u002Fv1` template around the native\n  `voice-benchmark-evidence\u002Fv1` file so workload, plan, identity, raw evidence,\n  friction, restoration, and decision roles remain explicit.\n  Use that contract's publication-ready result card and publication summary when the retained\n  voice artifact is meaningful; a format-only refresh does not rerun the\n  bounded voice benchmark.\n- Multi-sample STT qualification:\n  use `anvil-serving voice corpus prepare|validate` and\n  `anvil-serving voice benchmark --scope stt` through the canonical\n  `anvil-serving-stt-benchmark` skill. Do not collapse its primary human set,\n  separate synthetic set, repeated schedule, concurrency lane, or restoration\n  checks into the bounded one-turn benchmark above.\n\n## Result Packet\n\nReturn an `operator-workflow\u002Fv1` packet when coordinating with the broader\nworkbench. Use `artifacts` entries with `kind: \"voice-benchmark\"` or\n`kind: \"voice-sidecar-render\"`, `evidence_scope: \"voice-pipeline\"`, and\n`promotion_quality_evidence: false`. Keep `advisory_priors` empty unless\nexternal voice-specific priors are explicitly identified as advisory-only. Keep\n`promoted=false` and `human_gate_required=true` for any workflow that asks for\na capability-alias change, public binds, cloud enablement, or long-running service\nstarts.\n\nIf the existing `voice` and `voice sidecar` verbs cannot cover the request,\nreport the missing wrapper as a product gap after showing the closest safe\nexisting command and why it is insufficient.\n","# Anvil Serving Voice Ops\n\nUse this skill for voice-pipeline operations: validating sidecar manifests,\nrendering speech-to-speech commands, managing topology-owned STT\u002FTTS, managing the\npersistent Mini Realtime proxy, selecting profiles, running the Mini-local\nforwarding bridge, and collecting bounded voice benchmarks.\n\n## Start Here\n\n1. Read `README.md`, `CLAUDE.md`, and\n   `docs\u002FOPERATOR-SKILLS-AND-SUBAGENTS.md` before changing behavior or\n   reporting voice results as product evidence.\n2. Prefer `voice_manage` for structured STT\u002FTTS lifecycle and\n   `voice_proxy_manage` for persistent Realtime proxy lifecycle. Use the\n   existing CLI verbs for foreground, profile, sidecar, bridge, and benchmark\n   operations that do not have MCP wrappers:\n   `anvil-serving voice sidecar validate`,\n   `anvil-serving voice sidecar command`,\n   `anvil-serving voice sidecar compose`,\n   `anvil-serving voice audio {up|down|status|logs}`,\n   `anvil-serving voice profiles {list|validate}`,\n   `anvil-serving voice proxy {run|up|down|restart|status|logs|bridge}`, and\n   `anvil-serving voice benchmark`.\n3. Use `127.0.0.1` for local URLs. Do not introduce `localhost`.\n4. Pass secrets by environment variable name only. Do not put literal API keys\n   in manifests, commands, compose snippets, packets, or logs.\n5. Record the command host before interpreting loopback. In the reference\n   OpenClaw topology, Companion Node runs Gateway plus Anvil Voice Realtime\u002Fproxy\n   and reserves its 16 GB RAM for OpenClaw, Claude Code, and Codex. Do not run\n   STT\u002FTTS\u002FLLM model serves on Mini for reference testing. Use `dark-audio` for\n   Dark bridge ports or `mini-dark-audio-proxy` for Mini proxy ports forwarding\n   to Dark. `mini-audio` is an explicit optional same-host\u002Flocal-audio mode\n   only. A non-gateway checkout cannot validate Mini-local proxy paths by\n   calling its own `127.0.0.1`.\n6. Treat voice benchmark output as voice-pipeline evidence. Voice results do\n   not qualify an LLM serve or change a capability alias.\n7. Route multi-sample STT corpus qualification to\n   `skills\u002Fanvil-serving-stt-benchmark\u002FSKILL.md`. Keep this skill for\n   single-turn audio\u002Fvoice-pipeline evidence and topology lifecycle.\n8. Do not introduce one-off lifecycle, proxy, or port-forwarding scripts as the\n   operational path. If repeatable voice operation needs a new control, it\n   belongs in `anvil-serving voice` and, when agent-operated, MCP\u002Fcontroller.\n\n## Gates\n\nStop for a human gate before public or non-loopback Realtime binds, exposing a\nRealtime token, changing\ncloud\u002FSTT\u002FTTS spend, installing or updating external speech-to-speech packages,\nor starting long-running foreground services on a shared host.\n\n`voice sidecar command --with-auth` and `voice sidecar compose --with-auth`\nrender environment variable references for authenticated router calls. The\ncommand path returns argv, so explicitly warn that the env-var reference appears\nin process argv. The compose path includes an auth exposure comment in the\nrendered service. Use either path only on private hosts where process and\nDocker metadata are protected.\n\nNever use `anvil-serving voice benchmark` as a shortcut for LLM serve\nqualification. Voice benchmarks measure the voice pipeline: turn latency,\nTTFA, audio\u002FSTT\u002FTTS behavior, realtime transport, and user-perceived loop\nquality. They can support a voice-pipeline status report, but an LLM serve\nstill requires its own preflight, benchmark evidence, independent review, and\nhuman approval before an operator changes a capability alias.\n\nUse `workflow_packet_validate` before handing a voice packet to the broader\nworkbench. Voice artifacts must remain explicitly marked as non-promotion\nevidence.\n\n## Playbooks\n\n- Sidecar manifest validation:\n  `anvil-serving voice sidecar validate --config \u003Cmanifest>`.\n- Host command rendering:\n  `anvil-serving voice sidecar command --config \u003Cmanifest> --json`; add\n  `--with-auth` only when the operator accepts the argv exposure warning.\n- Compose rendering:\n  `anvil-serving voice sidecar compose --config \u003Cmanifest>`; inspect loopback\n  ports, `host.docker.internal` routing, image name, and auth comments.\n- Managed audio serves:\n  prefer `voice_manage` for agent\u002Fcontroller operation. The CLI equivalents are\n  `anvil-serving voice audio up --topology \u003Ctopology> --config \u003Cmanifest>` and\n  `voice audio down`; preview mutations first and require `--confirm` for live\n  execution. Use `anvil-serving voice audio status` and bounded\n  `anvil-serving voice audio logs` for reads.\n  External lifecycle serves\n  should be reported as skipped rather than forced into local Docker control.\n- Profile selection:\n  `anvil-serving voice profiles list --config \u003Cmanifest>` to list profiles,\n  `anvil-serving voice profiles validate --config \u003Cmanifest> --profile \u003Cname>`\n  before live work, then add `--profile \u003Cname>` to audio, proxy, or\n  `voice benchmark` when switching audio topology. For reference OpenClaw Talk\n  and candidate benchmarks, keep Mini model-free and select Dark-host audio or\n  a Mini-side proxy to Dark. Use `mini-audio` only when explicitly testing the\n  optional same-host\u002Flocal-audio mode.\n- Mini-to-Dark audio bridge:\n  `anvil-serving voice proxy bridge --topology \u003Ctopology> --config \u003Cmanifest>\n  --profile mini-dark-audio-proxy --dry-run` first. Run it on Mini. Require\n  topology-derived Dark targets and reject every non-loopback Mini listener.\n- Realtime proxy:\n  prefer `voice_proxy_manage` for the persistent Mini process. The CLI\n  equivalents are `anvil-serving voice proxy up` for persistent operation and\n  `anvil-serving voice proxy run` for a foreground diagnostic. Use\n  `down`\u002F`restart` only after preview and explicit\n  confirmation; use bounded `status`\u002F`logs` for inspection. Treat unreachable\n  STT\u002FTTS\u002Frouter endpoints as blockers. Never start audio lifecycle implicitly.\n- Bounded benchmark:\n  `anvil-serving voice benchmark --config \u003Cmanifest>` and capture JSON as\n  `voice-pipeline` evidence with `promoted=false`. The benchmark is an endpoint\n  client, not a Realtime-proxy lifecycle operation, so it has no proxy owner.\n  Use `--scope audio` for an LLM-free TTS-to-STT loop. Keep audio\n  topology and LLM selection separate: `--profile` selects the audio path and\n  `--candidate-overlay \u003Ctoml>` selects a candidate LLM configuration. Do not\n  combine `--candidate-overlay` with direct candidate flags. A voice benchmark\n  does not replace `anvil-serving eval preflight`, capacity benchmarking, or\n  repeated `anvil-serving eval benchmark quality` evidence for LLM serve\n  qualification.\n  Publish a meaningful retained result through\n  `skills\u002Fanvil-serving-benchmark-docs\u002FSKILL.md`; keep measured hardware\n  separate from protected\u002Fco-resident hardware and retain `promoted=false`.\n  Apply its `references\u002Fartifact-set-contract.md` and\n  `anvil-serving.benchmark-artifact-set\u002Fv1` template around the native\n  `voice-benchmark-evidence\u002Fv1` file so workload, plan, identity, raw evidence,\n  friction, restoration, and decision roles remain explicit.\n  Use that contract's publication-ready result card and publication summary when the retained\n  voice artifact is meaningful; a format-only refresh does not rerun the\n  bounded voice benchmark.\n- Multi-sample STT qualification:\n  use `anvil-serving voice corpus prepare|validate` and\n  `anvil-serving voice benchmark --scope stt` through the canonical\n  `anvil-serving-stt-benchmark` skill. Do not collapse its primary human set,\n  separate synthetic set, repeated schedule, concurrency lane, or restoration\n  checks into the bounded one-turn benchmark above.\n\n## Result Packet\n\nReturn an `operator-workflow\u002Fv1` packet when coordinating with the broader\nworkbench. Use `artifacts` entries with `kind: \"voice-benchmark\"` or\n`kind: \"voice-sidecar-render\"`, `evidence_scope: \"voice-pipeline\"`, and\n`promotion_quality_evidence: false`. Keep `advisory_priors` empty unless\nexternal voice-specific priors are explicitly identified as advisory-only. Keep\n`promoted=false` and `human_gate_required=true` for any workflow that asks for\na capability-alias change, public binds, cloud enablement, or long-running service\nstarts.\n\nIf the existing `voice` and `voice sidecar` verbs cannot cover the request,\nreport the missing wrapper as a product gap after showing the closest safe\nexisting command and why it is insufficient.\n",{"name":5,"description":7},"main","c02ba7c047f81b9e5fcc2a26434f6b5bfb82f5b0","skills\u002Fanvil-serving-voice-ops",0,"https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving","https:\u002F\u002Fgithub.com\u002Ffakoli","https:\u002F\u002Fsekoudoumbouya.com","2026-11-14 02:44:30",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add fakoli\u002Fanvil-serving --skill anvil-serving-voice-ops",{"tool":38,"command":39},"bunx","bunx skills add fakoli\u002Fanvil-serving --skill anvil-serving-voice-ops",{"tool":41,"command":42},"pnpm","pnpm dlx skills add fakoli\u002Fanvil-serving --skill anvil-serving-voice-ops","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Fskills\u002Fanvil-serving-voice-ops\u002FSKILL.md",true,[],[47,53,59,64,69,76],{"skillNo":48,"name":49,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M77RK15F1RY23K6YY54W7YQC","voice","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcli_reference_audit\u002Finput\u002Fskills\u002Fvoice\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":19},"01M77RK06RC7DDZQE1MEZHA9NQ","workbench","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcli_reference_audit\u002Finput\u002Fexamples\u002Fopenclaw\u002Fskills\u002Fworkbench\u002FSKILL.md",[],[],{"skillNo":60,"name":55,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M77RJY4CSSN04B1MRGPMSFH8","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcli_reference_audit\u002Finput\u002F.claude\u002Fskills\u002Fworkbench\u002FSKILL.md",[],[],{"skillNo":65,"name":55,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M77RJX2C6VRHYS7F452WTFKX","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcli_reference_audit\u002Finput\u002F.agents\u002Fskills\u002Fworkbench\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M77RJP56J64R6Y73WZCDJ03D","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\u002FCER and latency testing, candidate overlays, concurrency tests, restoration checks, and fail-closed stt-benchmark-evidence\u002Fv1 artifacts. Do not use for a single voice-pipeline turn or LLM qualification.","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Fskills\u002Fanvil-serving-stt-benchmark\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M77RJJB3P4FQGWDY306Q0KFB","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.","https:\u002F\u002Fgithub.com\u002Ffakoli\u002Fanvil-serving\u002Fblob\u002Fmain\u002Fskills\u002Fanvil-serving-release-readiness\u002FSKILL.md",[],[]]