[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7BGYWESQ44Z80VBZ3B3WB30":3,"skill-same:01M7BGYWESQ44Z80VBZ3B3WB30":85},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":13,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7BGYWESQ44Z80VBZ3B3WB30","nemo-relay-integrate-upstream",null,"Use this skill when assessing, extending, or implementing NeMo Relay support in an agent harness or agent framework, including coding agents and orchestration runtimes, when the host lacks Relay support or an existing integration needs deeper coverage. It identifies the host's execution boundaries and extension points, selects an appropriate attachment method for each boundary, and verifies the resulting coverage.","dawsonblock\u002FNEMO","dawsonblock","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F56236740?v=4","01M72Y87EG0A5JP9CKS7EYX8VM","01M7BGVFAXJ69RPAJYQ1XB52MZ",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: nemo-relay-integrate-upstream\ndescription: Use this skill when assessing, extending, or implementing NeMo Relay support in an agent harness or agent framework, including coding agents and orchestration runtimes, when the host lacks Relay support or an existing integration needs deeper coverage. It identifies the host's execution boundaries and extension points, selects an appropriate attachment method for each boundary, and verifies the resulting coverage.\nlicense: Apache-2.0\nmetadata:\n  author: NVIDIA Corporation and Affiliates\n---\n\n# Integrate Relay Into An Agent Harness\n\nUse this skill to determine the right NeMo Relay integration for an agent\nframework or harness that lacks support or needs deeper coverage. First\nidentify how the host runs tools, model calls, streams, subagents, and lifecycle\nevents and what its extension points can control. Then choose the integration\ndepth and Relay attachment methods that meet the required coverage and fit the\nhost's architecture, support model, and lifecycle, using an extension, separate\nadapter, core wiring, or hybrid as appropriate.\n\n## Choose The Work Mode\n\nChoose one mode from the user's request. Do not silently move from assessment\nto implementation.\n\n1. **Assess**: Determine whether and how Relay can integrate. Inspect only; do\n   not edit the host, Relay, issues, or pull requests. Read\n   [Assess The Host](references\u002Fassess-host.md).\n2. **Implement**: Make an already approved integration change. Reconfirm the\n   exact boundary and read [Implement The Integration](references\u002Fimplement-integration.md).\n3. **Qualify**: Test an existing integration at an exact host and Relay\n   revision. Treat this as read-only unless the user also requests fixes. Read\n   [Qualify The Integration](references\u002Fqualify-integration.md).\n\nIf the host already has a maintained Relay integration, treat it as the\nstarting point and verify its coverage against the requested outcome. Use it\nwhen it owns the required boundaries. When it does not, identify the missing\nownership and evaluate whether to extend the existing integration, combine it\nwith another attachment method, or move a named boundary deeper into host core.\nDo not create a competing adapter without first proving that gap. Moving from a\nhost plugin to core may be appropriate when requirements need deeper execution\nownership or when the intended first-party support, lifecycle, performance, and\nrelease model call for core integration. Hermes uses core to own real callbacks\nand lifecycle rather than relying on its former observability plugin.\n\nIf an application directly owns only a few tool or model callbacks, hand off to\n`nemo-relay-instrument-calls`.\n\n## Establish The Evidence Baseline\n\nBefore recommending a shape:\n\n1. Pin the host: record its exact revision or release and every distribution\n   form the integration must support.\n2. Establish the Relay release baseline separately. Check the latest stable and\n   prerelease GitHub releases, then read the matching Relay release notes,\n   support matrix, migration guide, package metadata, and relevant API docs.\n   Distinguish a published stable release from a release candidate, nightly or\n   alpha artifact, and current `main`. Choose and record one Relay target with\n   a reason; do not silently design against whichever source tree is open.\n3. Build a host-to-Relay packaging matrix. For each host distribution, record\n   its runtime and extension language, OS, architecture, libc, and package or\n   bundle format. Compare it with the Python, Node.js, Rust, gateway, or\n   advanced FFI or sidecar artifacts actually published for the selected Relay\n   release. Verify compatible versions and test installation and loading in\n   the real host distribution; source support or a CI build is not proof that\n   a compatible artifact was published or can load in a bundled host.\n4. Determine which host integration shapes exist: built-in Relay support, an\n   optional plugin or extension system, public middleware or adapter APIs,\n   direct core wiring, and configurable provider base URLs.\n5. Inspect those surfaces in source and tests. Trace representative tool and\n   model calls through retries, streaming, errors, cancellation, and subagents;\n   classify whether each surface owns a callback, honors a decision or\n   mutation, exposes paired lifecycle events, or only observes.\n6. Map every required boundary to an exact public API in the selected Relay\n   release. Then verify configuration, activation, health, cleanup, and\n   duplicate-instrumentation behavior. Binding availability alone does not\n   prove that the required API or semantics exist, and an integration that can\n   be silently skipped needs an explicit verification story.\n\nUse source and targeted tests as evidence. When only a closed binary or public\ndocumentation is available, state that limitation and do not claim complete\ncoverage.\n\n## Build An Ownership Matrix\n\nEvaluate these surfaces independently:\n\n- process, session, run, turn, and compaction lifecycle;\n- tool request, execution, result, and MCP bridge behavior;\n- LLM request, provider attempt, response, and streaming finalization;\n- subagent creation and cross-task, thread, or process context propagation;\n- approvals, human-in-the-loop decisions, retries, and queued work;\n- configuration, activation, health, flush, and shutdown.\n\nSplit a broad surface into separate rows when its paths have different owners\nor hooks. For example, do not combine a main-loop model call, provider retry,\ncompaction call, and extension-owned provider call into one LLM row.\n\nFor each independently owned path, record:\n\n- the exact call path, Relay attachment point, covered dispatch families, and\n  bypasses;\n- logical-versus-physical cardinality across retries, streams, re-entry, and\n  concurrency;\n- the owning scope stack or execution context, callback thread or event loop,\n  and whether siblings can finish out of order;\n- callback or network ownership, hook ordering, honored returns, and whether\n  the host can block, mutate, replace, or only observe;\n- stable identity, parent context, JSON projection, and streaming behavior;\n- success, failure, cancellation, retry, stream, and shutdown closure\n  ownership, including which process component owns Relay activation and\n  final drain.\n\nDo not summarize a host as simply \"hooked\" or \"not hooked.\" One useful hook\ndoes not prove the other surfaces or every dispatch path within that surface.\n\n## Classify The Host Attachment Contract\n\nRead [Host Attachment And Hook Patterns](references\u002Fhost-attachment-patterns.md).\nCore adapters and wrapper middleware can support managed execution only when\nthey receive the real callback. Awaited prehooks can support blocking or\nmutation only when the host applies their return value. Paired events support\nmanual lifecycle observation, while session and subagent events support causal\nstructure. A compatible base URL covers provider traffic only. Plugin lifecycle\nhooks must make activation, health, drain, and cleanup visible.\n\n## Audit Concurrency And Lifecycle\n\nRead [Concurrency And Lifecycle](references\u002Fconcurrency-and-lifecycle.md) when\nthe host can overlap requests, turns, tools, model attempts, streams,\nsubagents, callbacks, or shutdown. Determine whether the integration controls\ntask creation, receives only paired callbacks after scheduling, crosses threads\nor event loops, or shares process-global Relay state across multiple host\nobjects.\n\nA scope handle establishes identity and parentage; it is not the execution\ncontext that owns the mutable scope stack or makes scope-local middleware\nvisible. Give every managed call one adapter owner, preserve the callback and\nstream lifetime, and keep host-instance ownership separate from live Relay\noperations and process-global plugin activation.\n\n## Choose The Host Integration Depth\n\nUse the evidence and ownership matrix to decide where Relay should attach.\nChoose based on required coverage, host architecture, adoption model, support\nand release ownership, operational lifecycle, and performance.\n\n| Host integration shape | Choose it when | Main trade-off |\n|---|---|---|\n| Existing maintained integration | Relay already supports the host and the integration covers the required boundaries. | Do not create a competing adapter or duplicate instrumentation. |\n| Optional host plugin or extension | A public extension API exposes enough awaited, mutable, and paired hooks for the required behavior. | Easy to adopt and remove, but limited to the ownership the host exposes. |\n| Separate integration package | Public framework callbacks are sufficient, but the host has no suitable plugin registry or should not depend directly on Relay. | Keeps release ownership separate, but compatibility must be maintained across both projects. |\n| Direct host-core integration | Required tool, model, stream, subagent, or lifecycle ownership is unavailable to extensions and Relay is intended to be a first-party capability. | Provides the deepest coverage but couples Relay to the host's runtime and release lifecycle. |\n| Hybrid | Different surfaces have different owners, such as a host extension for tools and the Relay gateway for model traffic. | Often the honest shape, but activation, correlation, and duplicate instrumentation require explicit design. |\n\nDo not confuse a host plugin with a Relay plugin. A host plugin or extension\nattaches Relay to the harness. Relay plugins configure middleware, policy, and\nexporters after the host has attached Relay to its execution boundaries.\n\nDo not treat an extension or core integration as the default. Justify the\nchosen depth from named requirements and the intended product relationship.\n\n## Select A Relay Method Per Surface\n\n| Host boundary | Preferred Relay method | Honest capability |\n|---|---|---|\n| Host yields the actual tool or LLM callback | Managed execution wrapper | Relay owns the full middleware and lifecycle boundary. |\n| Host exposes paired start\u002Fend events only | Explicit lifecycle APIs | Observation and correlation; middleware does not run automatically. |\n| Host awaits an allow-or-block prehook | Standalone conditional execution | Policy can reject before the host continues. |\n| Host honors a mutable request prehook | Standalone request intercepts | Relay can return a rewritten request for the host to apply. |\n| Host supports an OpenAI- or Anthropic-compatible base URL | Relay gateway | Managed provider traffic only; this does not cover tools or host lifecycle. |\n| Host exposes a milestone without a call boundary | Mark | A correlated event, not managed execution. |\n| Work crosses a task, thread, or process boundary | Explicit context propagation | Parentage only when the receiving side restores the context correctly. |\n\nChoose these independently. For example, an extension may gate tools, replay\nlifecycle events, and route model HTTP traffic through the gateway. Do not call\nthat a single managed integration.\n\n## Keep Capability Claims Separate\n\nUse these terms precisely:\n\n- **Observable**: Relay receives enough data to emit an event.\n- **Correlated**: Events have stable parent, call, and run identity.\n- **Evaluable**: Relay can compute a policy decision.\n- **Enforceable**: The host honors a reject or block before execution.\n- **Mutable**: The host applies Relay's returned request or result.\n- **Managed**: Relay owns the callback and executes its complete managed\n  middleware and lifecycle sequence.\n- **Complete**: Every in-scope path is covered, including retries, streams,\n  subagents, errors, and alternate providers.\n\nNever infer enforcement from evaluation, mutation from observation, managed\nbehavior from paired events, or tool coverage from a provider gateway. A tool\nhook sees the declared tool call, not necessarily nested commands or other work\nperformed inside the tool.\n\n## Produce A Decision Record\n\nFor an assessment or review, report:\n\n1. exact revisions and distribution targets;\n2. the ownership matrix with source or test evidence;\n3. the recommended Relay method and proven capability for each surface;\n4. known gaps and escape paths;\n5. the required upstream, Relay, and deployment changes and their owners;\n6. a deterministic qualification plan and unresolved ownership decisions.\n\nSeparate facts proven at the exact revision from hypotheses that require a\nspike. Do not turn every missing capability into a Relay API proposal. When the\nmissing ownership belongs to the host, determine whether a stable extension\npoint or first-party core change should provide it.\n\n## Handoffs\n\n- First-time Relay setup -> `nemo-relay-install`, then\n  `nemo-relay-get-started`.\n- A few application-owned callbacks -> `nemo-relay-instrument-calls`.\n- Scope leakage across concurrent work ->\n  `nemo-relay-instrument-context-isolation`.\n- Provider or typed value projection ->\n  `nemo-relay-instrument-typed-wrappers`.\n- Reusable configuration-driven Relay behavior -> `nemo-relay-plugin-build`.\n- Missing events or activation failures ->\n  `nemo-relay-debug-runtime-integration`.\n","# Integrate Relay Into An Agent Harness\n\nUse this skill to determine the right NeMo Relay integration for an agent\nframework or harness that lacks support or needs deeper coverage. First\nidentify how the host runs tools, model calls, streams, subagents, and lifecycle\nevents and what its extension points can control. Then choose the integration\ndepth and Relay attachment methods that meet the required coverage and fit the\nhost's architecture, support model, and lifecycle, using an extension, separate\nadapter, core wiring, or hybrid as appropriate.\n\n## Choose The Work Mode\n\nChoose one mode from the user's request. Do not silently move from assessment\nto implementation.\n\n1. **Assess**: Determine whether and how Relay can integrate. Inspect only; do\n   not edit the host, Relay, issues, or pull requests. Read\n   [Assess The Host](references\u002Fassess-host.md).\n2. **Implement**: Make an already approved integration change. Reconfirm the\n   exact boundary and read [Implement The Integration](references\u002Fimplement-integration.md).\n3. **Qualify**: Test an existing integration at an exact host and Relay\n   revision. Treat this as read-only unless the user also requests fixes. Read\n   [Qualify The Integration](references\u002Fqualify-integration.md).\n\nIf the host already has a maintained Relay integration, treat it as the\nstarting point and verify its coverage against the requested outcome. Use it\nwhen it owns the required boundaries. When it does not, identify the missing\nownership and evaluate whether to extend the existing integration, combine it\nwith another attachment method, or move a named boundary deeper into host core.\nDo not create a competing adapter without first proving that gap. Moving from a\nhost plugin to core may be appropriate when requirements need deeper execution\nownership or when the intended first-party support, lifecycle, performance, and\nrelease model call for core integration. Hermes uses core to own real callbacks\nand lifecycle rather than relying on its former observability plugin.\n\nIf an application directly owns only a few tool or model callbacks, hand off to\n`nemo-relay-instrument-calls`.\n\n## Establish The Evidence Baseline\n\nBefore recommending a shape:\n\n1. Pin the host: record its exact revision or release and every distribution\n   form the integration must support.\n2. Establish the Relay release baseline separately. Check the latest stable and\n   prerelease GitHub releases, then read the matching Relay release notes,\n   support matrix, migration guide, package metadata, and relevant API docs.\n   Distinguish a published stable release from a release candidate, nightly or\n   alpha artifact, and current `main`. Choose and record one Relay target with\n   a reason; do not silently design against whichever source tree is open.\n3. Build a host-to-Relay packaging matrix. For each host distribution, record\n   its runtime and extension language, OS, architecture, libc, and package or\n   bundle format. Compare it with the Python, Node.js, Rust, gateway, or\n   advanced FFI or sidecar artifacts actually published for the selected Relay\n   release. Verify compatible versions and test installation and loading in\n   the real host distribution; source support or a CI build is not proof that\n   a compatible artifact was published or can load in a bundled host.\n4. Determine which host integration shapes exist: built-in Relay support, an\n   optional plugin or extension system, public middleware or adapter APIs,\n   direct core wiring, and configurable provider base URLs.\n5. Inspect those surfaces in source and tests. Trace representative tool and\n   model calls through retries, streaming, errors, cancellation, and subagents;\n   classify whether each surface owns a callback, honors a decision or\n   mutation, exposes paired lifecycle events, or only observes.\n6. Map every required boundary to an exact public API in the selected Relay\n   release. Then verify configuration, activation, health, cleanup, and\n   duplicate-instrumentation behavior. Binding availability alone does not\n   prove that the required API or semantics exist, and an integration that can\n   be silently skipped needs an explicit verification story.\n\nUse source and targeted tests as evidence. When only a closed binary or public\ndocumentation is available, state that limitation and do not claim complete\ncoverage.\n\n## Build An Ownership Matrix\n\nEvaluate these surfaces independently:\n\n- process, session, run, turn, and compaction lifecycle;\n- tool request, execution, result, and MCP bridge behavior;\n- LLM request, provider attempt, response, and streaming finalization;\n- subagent creation and cross-task, thread, or process context propagation;\n- approvals, human-in-the-loop decisions, retries, and queued work;\n- configuration, activation, health, flush, and shutdown.\n\nSplit a broad surface into separate rows when its paths have different owners\nor hooks. For example, do not combine a main-loop model call, provider retry,\ncompaction call, and extension-owned provider call into one LLM row.\n\nFor each independently owned path, record:\n\n- the exact call path, Relay attachment point, covered dispatch families, and\n  bypasses;\n- logical-versus-physical cardinality across retries, streams, re-entry, and\n  concurrency;\n- the owning scope stack or execution context, callback thread or event loop,\n  and whether siblings can finish out of order;\n- callback or network ownership, hook ordering, honored returns, and whether\n  the host can block, mutate, replace, or only observe;\n- stable identity, parent context, JSON projection, and streaming behavior;\n- success, failure, cancellation, retry, stream, and shutdown closure\n  ownership, including which process component owns Relay activation and\n  final drain.\n\nDo not summarize a host as simply \"hooked\" or \"not hooked.\" One useful hook\ndoes not prove the other surfaces or every dispatch path within that surface.\n\n## Classify The Host Attachment Contract\n\nRead [Host Attachment And Hook Patterns](references\u002Fhost-attachment-patterns.md).\nCore adapters and wrapper middleware can support managed execution only when\nthey receive the real callback. Awaited prehooks can support blocking or\nmutation only when the host applies their return value. Paired events support\nmanual lifecycle observation, while session and subagent events support causal\nstructure. A compatible base URL covers provider traffic only. Plugin lifecycle\nhooks must make activation, health, drain, and cleanup visible.\n\n## Audit Concurrency And Lifecycle\n\nRead [Concurrency And Lifecycle](references\u002Fconcurrency-and-lifecycle.md) when\nthe host can overlap requests, turns, tools, model attempts, streams,\nsubagents, callbacks, or shutdown. Determine whether the integration controls\ntask creation, receives only paired callbacks after scheduling, crosses threads\nor event loops, or shares process-global Relay state across multiple host\nobjects.\n\nA scope handle establishes identity and parentage; it is not the execution\ncontext that owns the mutable scope stack or makes scope-local middleware\nvisible. Give every managed call one adapter owner, preserve the callback and\nstream lifetime, and keep host-instance ownership separate from live Relay\noperations and process-global plugin activation.\n\n## Choose The Host Integration Depth\n\nUse the evidence and ownership matrix to decide where Relay should attach.\nChoose based on required coverage, host architecture, adoption model, support\nand release ownership, operational lifecycle, and performance.\n\n| Host integration shape | Choose it when | Main trade-off |\n|---|---|---|\n| Existing maintained integration | Relay already supports the host and the integration covers the required boundaries. | Do not create a competing adapter or duplicate instrumentation. |\n| Optional host plugin or extension | A public extension API exposes enough awaited, mutable, and paired hooks for the required behavior. | Easy to adopt and remove, but limited to the ownership the host exposes. |\n| Separate integration package | Public framework callbacks are sufficient, but the host has no suitable plugin registry or should not depend directly on Relay. | Keeps release ownership separate, but compatibility must be maintained across both projects. |\n| Direct host-core integration | Required tool, model, stream, subagent, or lifecycle ownership is unavailable to extensions and Relay is intended to be a first-party capability. | Provides the deepest coverage but couples Relay to the host's runtime and release lifecycle. |\n| Hybrid | Different surfaces have different owners, such as a host extension for tools and the Relay gateway for model traffic. | Often the honest shape, but activation, correlation, and duplicate instrumentation require explicit design. |\n\nDo not confuse a host plugin with a Relay plugin. A host plugin or extension\nattaches Relay to the harness. Relay plugins configure middleware, policy, and\nexporters after the host has attached Relay to its execution boundaries.\n\nDo not treat an extension or core integration as the default. Justify the\nchosen depth from named requirements and the intended product relationship.\n\n## Select A Relay Method Per Surface\n\n| Host boundary | Preferred Relay method | Honest capability |\n|---|---|---|\n| Host yields the actual tool or LLM callback | Managed execution wrapper | Relay owns the full middleware and lifecycle boundary. |\n| Host exposes paired start\u002Fend events only | Explicit lifecycle APIs | Observation and correlation; middleware does not run automatically. |\n| Host awaits an allow-or-block prehook | Standalone conditional execution | Policy can reject before the host continues. |\n| Host honors a mutable request prehook | Standalone request intercepts | Relay can return a rewritten request for the host to apply. |\n| Host supports an OpenAI- or Anthropic-compatible base URL | Relay gateway | Managed provider traffic only; this does not cover tools or host lifecycle. |\n| Host exposes a milestone without a call boundary | Mark | A correlated event, not managed execution. |\n| Work crosses a task, thread, or process boundary | Explicit context propagation | Parentage only when the receiving side restores the context correctly. |\n\nChoose these independently. For example, an extension may gate tools, replay\nlifecycle events, and route model HTTP traffic through the gateway. Do not call\nthat a single managed integration.\n\n## Keep Capability Claims Separate\n\nUse these terms precisely:\n\n- **Observable**: Relay receives enough data to emit an event.\n- **Correlated**: Events have stable parent, call, and run identity.\n- **Evaluable**: Relay can compute a policy decision.\n- **Enforceable**: The host honors a reject or block before execution.\n- **Mutable**: The host applies Relay's returned request or result.\n- **Managed**: Relay owns the callback and executes its complete managed\n  middleware and lifecycle sequence.\n- **Complete**: Every in-scope path is covered, including retries, streams,\n  subagents, errors, and alternate providers.\n\nNever infer enforcement from evaluation, mutation from observation, managed\nbehavior from paired events, or tool coverage from a provider gateway. A tool\nhook sees the declared tool call, not necessarily nested commands or other work\nperformed inside the tool.\n\n## Produce A Decision Record\n\nFor an assessment or review, report:\n\n1. exact revisions and distribution targets;\n2. the ownership matrix with source or test evidence;\n3. the recommended Relay method and proven capability for each surface;\n4. known gaps and escape paths;\n5. the required upstream, Relay, and deployment changes and their owners;\n6. a deterministic qualification plan and unresolved ownership decisions.\n\nSeparate facts proven at the exact revision from hypotheses that require a\nspike. Do not turn every missing capability into a Relay API proposal. When the\nmissing ownership belongs to the host, determine whether a stable extension\npoint or first-party core change should provide it.\n\n## Handoffs\n\n- First-time Relay setup -> `nemo-relay-install`, then\n  `nemo-relay-get-started`.\n- A few application-owned callbacks -> `nemo-relay-instrument-calls`.\n- Scope leakage across concurrent work ->\n  `nemo-relay-instrument-context-isolation`.\n- Provider or typed value projection ->\n  `nemo-relay-instrument-typed-wrappers`.\n- Reusable configuration-driven Relay behavior -> `nemo-relay-plugin-build`.\n- Missing events or activation failures ->\n  `nemo-relay-debug-runtime-integration`.\n",{"name":5,"license":14,"metadata":23,"description":7},{"author":24},"NVIDIA Corporation and Affiliates","main","13944d7d2b1a4bc9aa91fa061bed6fe63fad7979","skills\u002Fnemo-relay-integrate-upstream","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO","https:\u002F\u002Fgithub.com\u002Fdawsonblock","2026-09-30 06:03:35",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add dawsonblock\u002FNEMO --skill nemo-relay-integrate-upstream",{"tool":37,"command":38},"bunx","bunx skills add dawsonblock\u002FNEMO --skill nemo-relay-integrate-upstream",{"tool":40,"command":41},"pnpm","pnpm dlx skills add dawsonblock\u002FNEMO --skill nemo-relay-integrate-upstream","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002FSKILL.md",true,[45,51,55,59,63,67,71,75,79],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"BENCHMARK.md","REFERENCE","参考文档",5255,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002FBENCHMARK.md",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"evals\u002Fevals.json",15626,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Fevals\u002Fevals.json",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Fassess-host.md",9430,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Freferences\u002Fassess-host.md",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Fconcurrency-and-lifecycle.md",7320,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Freferences\u002Fconcurrency-and-lifecycle.md",{"path":64,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":65,"sourceHtmlUrl":66},"references\u002Fhost-attachment-patterns.md",11690,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Freferences\u002Fhost-attachment-patterns.md",{"path":68,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":69,"sourceHtmlUrl":70},"references\u002Fimplement-integration.md",7840,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Freferences\u002Fimplement-integration.md",{"path":72,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":73,"sourceHtmlUrl":74},"references\u002Fqualify-integration.md",5930,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Freferences\u002Fqualify-integration.md",{"path":76,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":77,"sourceHtmlUrl":78},"skill-card.md",4660,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Fskill-card.md",{"path":80,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":83,"sourceHtmlUrl":84},"skill.oms.sig","OTHER","其他",5853,"https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-integrate-upstream\u002Fskill.oms.sig",[86,93,100,107,114,121],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":18},"01M7BGZHFG0R2E7EN6JKK8A6XK","nemo-relay-plugin-observability","Use this skill when choosing or configuring NeMo Relay 0.6 or 0.7 observability through the built-in plugin, subscribers, or exporters, including raw ATOF events, ATIF trajectories, OpenTelemetry, OpenInference, or custom event handling.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-plugin-observability\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":18},"01M7BGZDJ0YSBESTS38D9FDRNA","nemo-relay-plugin-build","Use this skill when building or packaging reusable NeMo Relay runtime behavior as an embedded configuration component or a manifest-backed `rust_dynamic` native or `worker` gRPC plugin, with deterministic validation and rollback-safe registration.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-plugin-build\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":18},"01M7BGZ9KYD60WWHH6C5H9DKPM","nemo-relay-plugin-adaptive-tuning","Use this skill when baseline NeMo Relay instrumentation exists and the user wants to configure or evaluate adaptive plugin behavior, including telemetry, state, adaptive_hints, tool_parallelism, ACG, hint consumption, or measured rollout.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-plugin-adaptive-tuning\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":18},"01M7BGZ3XE0H92KR4E3ENNXQNX","nemo-relay-migrate-from-flow","Use this skill when migrating applications, examples, integrations, documentation, manifests, or repository code from NeMo Flow to NeMo Relay across Python, Rust, Node.js, Go, C FFI, CLI, configuration, and observability surfaces.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-migrate-from-flow\u002FSKILL.md",[],[],{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":118,"categories":119,"keywords":120,"publishedAt":18},"01M7BGYRQ1PTFVVC8KGA8ZE4MN","nemo-relay-instrument-typed-wrappers","Use this skill when adding NeMo Relay typed wrappers, domain types, or provider codecs while preserving JSON middleware semantics and caller-visible behavior.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-instrument-typed-wrappers\u002FSKILL.md",[],[],{"skillNo":122,"name":123,"nameZh":6,"description":124,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":127,"publishedAt":18},"01M7BGYMR22V7RZZA8VFJ4852B","nemo-relay-instrument-context-isolation","Use this skill when concurrent requests, async tasks, threads, workers, goroutines, or agents need independent NeMo Relay scope stacks and correct ancestry propagation.","https:\u002F\u002Fgithub.com\u002Fdawsonblock\u002FNEMO\u002Fblob\u002Fmain\u002Fskills\u002Fnemo-relay-instrument-context-isolation\u002FSKILL.md",[],[]]