[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7416EHCYH7YCJ9XPSQ2844V":3,"skill-same:01M7416EHCYH7YCJ9XPSQ2844V":56},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"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":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":28,"ownerHtmlUrl":29,"ownerBlogUrl":28,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7416EHCYH7YCJ9XPSQ2844V","network-handoff",null,"Use before generating, changing, or reviewing networking, IPC, or cross-process handoffs, including requests, retries, deadlines, reconnect, process launch, configuration apply, cancellation, and cleanup. Establish confirmation, identity, authority, reconciliation, and producer compatibility before implementation.","openclaw\u002Fopenclaw-windows-node","openclaw","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F252820863?v=4","01M698P3FSAZ6B4R928ZF19KPT","01M698Q6SND6RHZDPP8ZM0NSP7",2198,"MIT","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fnetwork-handoff\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: network-handoff\ndescription: Use before generating, changing, or reviewing networking, IPC, or cross-process handoffs, including requests, retries, deadlines, reconnect, process launch, configuration apply, cancellation, and cleanup. Establish confirmation, identity, authority, reconciliation, and producer compatibility before implementation.\n---\n\n# Network and handoff contracts\n\nFor code-generating and reviewing agents, across all networking and IPC, not\njust onboarding. The outcome is the smallest change whose next-step postcondition\nis established by the actual producer contract. Read-only work stays read-only;\nthis skill grants no runtime, installation, upstream-write, or publication authority.\n\n## Trace before editing\n\n1. Name the reported behavior and the next step it must safely enable. Follow\n   the real caller, transport, receiver, side effect, state owner, and tests.\n   Read the affected rows in [the ownership ledger](..\u002F..\u002F..\u002Fdocs\u002FARCHITECTURE.md)\n   and the owning guide linked from [AGENTS.md](..\u002F..\u002F..\u002FAGENTS.md).\n   Do not create another ownership ledger or a parallel client\u002Fstate store.\n2. Record worktree\u002Fbase\u002FHEAD and producer repository\u002Fsource revision. Separately\n   establish the installed producer build, negotiated protocol\u002Fcapabilities and\n   selected resource\u002Fprofile. Main-branch source is not installed-version proof.\n   Historical documents are hypotheses to recheck, not shipping specifications.\n3. Inventory existing request\u002Fresult schemas, status queries, event delivery,\n   persistence, authorization, retries, limits, cancellation and cleanup owners.\n   Read implementations and tests, not only method names or an exit code.\n4. Write the boundary invariants below in the task's plan or private evidence.\n   Mark unknown or inapplicable fields explicitly. Do not invent infrastructure\n   for fields the changed contract does not need.\n\n| Question | Required decision |\n|---|---|\n| What does the acknowledgment establish? | Received, durably accepted, executing, committed, or ready; define persistence\u002Ffailure scope |\n| What does the next step require? | Exact resource, generation, desired\u002Fapplied revision, freshness and authorization |\n| What identifies and authorizes work? | Separate request\u002Foperation ID, resource identity, runtime boot\u002Fconnection generation, revision, and authority\u002Ffencing |\n| Who knows the outcome? | Authoritative query and its consistency, retention, unknown\u002Fexpired\u002Fnot-found meanings; event recovery if used |\n| Who may retry or clean up? | One coordinating retry owner, mutation idempotency contract, receiver-side ownership check, cancellation\u002Fcompensation owner |\n| What bounds waiting and execution? | Per-call limit, remaining end-to-end budget, no-progress definition, backend lifetime and cleanup lifetime |\n\n## Choose the minimal sufficient protocol\n\nA bounded synchronous call is enough when it returns the required committed\nresult with documented semantics. An existing authoritative read plus bounded\npolling may be simpler than a watch. Long-running work should acknowledge\nacceptance with a queryable operation handle rather than hold the initial RPC\nuntil readiness. Distinguish volatile receipt from durable acceptance explicitly.\nDo not add streaming, a broker, a database or a general operation framework\nunless the boundary's requirements actually need it.\n\nApply the following invariants to the affected path:\n\n1. **Identity is not authority.** Correlate replies to their request, resource,\n   boot\u002Fconnection generation and revision as applicable. A UUID, PID, open port\n   or matching config is not permission. Preserve credential and provenance\n   checks. The receiver must atomically validate current ownership\u002Ffencing with\n   its side effect, through its existing serialization\u002Fconditional-write seam.\n   A client precheck alone leaves a race.\n2. **State establishes correctness; events accelerate observation.** Use an\n   authoritative query with adequate freshness\u002Fconsistency. Snapshot+watch needs\n   a producer-supported revision\u002Fcursor bridge, or subscribe\u002Fbuffer plus a\n   documented snapshot barrier. Recover by resnapshot on reconnect or expired\n   history; reject duplicate, reordered and stale-generation events. Never\n   assume arbitrary events have etcd's delivery guarantees. Without a safe\n   bridge use bounded polling if the query suffices.\n3. **Operation outcome and runtime health differ.** Terminal operation outcomes\n   are immutable for their documented retention lifetime. Observation sequence\n   increases within its declared scope, but health\u002Freadiness may regress.\n   A new boot is a new generation, even on the same endpoint. An old successful\n   launch does not prove that the current runtime is ready.\n4. **Timeout is an observation result.** Before dispatch, expiry may prevent\n   work. After dispatch, timeout\u002Fdisconnect may leave the outcome unknown,\n   including effects already committed. Keep that uncertainty separate from\n   authoritative terminal failure. Query by identity; distinguish history\n   expiry from proof of absence. Replay only under an applicable idempotency\n   contract. Never roll back merely because an acknowledgment was lost.\n5. **Bound every call and the whole wait.** Propagate the remaining end-to-end\n   duration, accounting for queueing, connection\u002Fauthentication, reads, polling\n   and backoff. Measure local elapsed time monotonically; do not compare\n   unsynchronized process clocks as a deadline protocol. Use existing named,\n   configurable repository limit patterns with measured\u002Fjustified values.\n   Responsiveness heartbeat is not useful progress. Only documented forward\n   progress renews a no-progress allowance; nothing renews the outer ceiling.\n   Client wait expiry does not implicitly terminate accepted backend work.\n   Specify backend execution, result retention and cleanup ownership separately.\n6. **Idempotency must be implemented, not inferred.** Establish caller\u002Fresource\n   scope, same-key\u002Fdifferent-parameters conflict, concurrent deduplication,\n   retention\u002Fexpiry and the crash-safe relationship of effect and result record.\n   If external effects cannot commit atomically with the record, require\n   recoverable intent\u002Freconciliation and name residual ambiguity. UUIDs and\n   in-memory caches do not establish durable exactly-once execution.\n7. **Retry at one coordinating layer.** Inventory SDK, transport and workflow\n   retries to avoid multiplication. Define bounded attempts\u002Frate, transient\n   classification, capped backoff with jitter and supported retry hints within\n   the remaining budget. Query retries and mutation retries are different;\n   authentication\u002Fconflict\u002Fprovenance failures are not generic transient errors.\n8. **Cancellation is not rollback.** Distinguish cancellation requested,\n   execution stopped, descendants drained and cleanup\u002Fcompensation complete.\n   Completion may win the race. Query the actual terminal outcome. Compensation\n   must be explicit, idempotent, resumable and scoped to current authority at\n   the mutation boundary. Never restore an old snapshot over concurrent edits.\n9. **Prove the next step, not a proxy.** Saved config is not applied config.\n   Alive, listener owned, startup ready, operator admitted and capability\n   approved are different predicates. Bind the required predicate to current\n   generation\u002Frevision\u002Ffreshness and the actual caller's authorization.\n   Current matching state can permit a next step when the contract says so;\n   it does not prove which historical operation caused it.\n\n## Inspect producer support and disposition gaps\n\nUse the existing supported contract, including capability\u002Fversion negotiation,\nbefore declaring a feature missing. Implement only consumer behavior justified\nby the supported producer versions. An honest compatibility path may offer a\nweaker explicitly named guarantee if the next step accepts it. Otherwise report\nthe precise unavailable confirmation and stop that transition, not the whole\nunrelated product.\n\nFor each genuine producer gap, record:\n\n- Repository, authoritative producer file\u002Fsymbol, pinned SHA and source\u002Ftest\n  links; distinguish implementation, documented guarantee, runtime proof and\n  installed-version unknowns.\n- Required missing contract, existing support checked, and why consumer reads,\n  local IDs or timing cannot establish it. Separate historical attribution from\n  present-state reconciliation.\n- Minimal versioned producer proposal: schema\u002Fsemantics, authorization\u002Ffencing,\n  persistence\u002Fretention, query\u002Frecovery and errors only as needed for this path.\n- Compatibility and rollout: capability discovery, producer-first deployment,\n  supported older versions, consumer gating and explicit blocked\u002Fdegraded state.\n  Do not guess a minimum released version from a source commit.\n- Acceptance tests covering the failure window that requires producer support.\n  Suggest the update to its owning repository; do not silently implement a\n  consumer workaround or mutate another repository without authority.\n\nThe [pinned onboarding case study](references\u002Fupstream-gaps.md) illustrates this\ndisposition, not a permanent capability catalog. Recheck it for the task's\nactual producer revision. [Research notes](references\u002Fprinciples.md) explain\nthe source-supported principles without importing vendor-specific requirements.\n\nNo sleeps, longer timeouts, log parsing, unconditional cleanup or speculative\nfallback may substitute for required producer guarantees. Diagnostic logs can\nexplain a failure, not become an undocumented protocol. Preserve necessary\ntimeouts and all credential\u002Fprovenance checks.\n\n## Select failure proof and hand off\n\nUse the existing test harness and the [test authoring](..\u002Ftest-audit\u002FSKILL.md),\n[test selection](..\u002Fopenclaw-testing\u002FSKILL.md) and\n[proof](..\u002Fopenclaw-proof-validation\u002FSKILL.md) routes. Select applicable cases,\nincluding receiver-side tests where a consumer mock cannot prove the guarantee:\n\n| Failure injection | Observable assertion |\n|---|---|\n| Drop acknowledgment after acceptance\u002Fcommit | Reconcile same identity; no blind mutation replay or false terminal failure |\n| Duplicate\u002Freorder events; disconnect at snapshot\u002Fsubscribe | No missed required transition or stale state regression; recover from reconnect\u002Fhistory expiry |\n| Crash between effect and receipt; concurrent same key | Recover intent\u002Foutcome without duplicate effect; reject changed parameters; expose expiry ambiguity |\n| Stale cleanup or delayed write after a new owner | Receiver rejects obsolete authority at the side-effect boundary; newer state survives |\n| Cancel as completion commits; child still running | Terminal winner preserved; requested\u002Fstopped\u002Fdrained\u002Fcompensated remain distinct |\n| New boot reuses PID\u002Fport; stale config\u002Fhealth reply | Reject stale generation\u002Frevision; verify fresh own-identity admission as required |\n| Exact timeout\u002Fno-progress boundaries with virtual clocks | Finite calls and outer ceiling; wall-clock jumps\u002Fheartbeats do not extend budget; pre\u002Fpost-dispatch outcomes differ |\n| Retry storm, unsupported producer, authorization failure | One bounded retry owner; no multiplied retries or invented compatibility success |\n\nFor a small read-only query change, retain only the relevant invariants\u002Fcases\nand explain exclusions; do not demand a full onboarding E2E suite. For runtime\nclaims, collect current-head evidence at the real changed boundary or name the\nblocker. A fixture proves only its simulated contract.\n\nHandoff: changed owner and invariant, exact source\u002Fbuild\u002Fversion evidence,\nconfirmation actually achieved, commands\u002Fresults, unverified boundaries and\nupstream dispositions. Do not claim stronger guarantees than the producer.\n","# Network and handoff contracts\n\nFor code-generating and reviewing agents, across all networking and IPC, not\njust onboarding. The outcome is the smallest change whose next-step postcondition\nis established by the actual producer contract. Read-only work stays read-only;\nthis skill grants no runtime, installation, upstream-write, or publication authority.\n\n## Trace before editing\n\n1. Name the reported behavior and the next step it must safely enable. Follow\n   the real caller, transport, receiver, side effect, state owner, and tests.\n   Read the affected rows in [the ownership ledger](..\u002F..\u002F..\u002Fdocs\u002FARCHITECTURE.md)\n   and the owning guide linked from [AGENTS.md](..\u002F..\u002F..\u002FAGENTS.md).\n   Do not create another ownership ledger or a parallel client\u002Fstate store.\n2. Record worktree\u002Fbase\u002FHEAD and producer repository\u002Fsource revision. Separately\n   establish the installed producer build, negotiated protocol\u002Fcapabilities and\n   selected resource\u002Fprofile. Main-branch source is not installed-version proof.\n   Historical documents are hypotheses to recheck, not shipping specifications.\n3. Inventory existing request\u002Fresult schemas, status queries, event delivery,\n   persistence, authorization, retries, limits, cancellation and cleanup owners.\n   Read implementations and tests, not only method names or an exit code.\n4. Write the boundary invariants below in the task's plan or private evidence.\n   Mark unknown or inapplicable fields explicitly. Do not invent infrastructure\n   for fields the changed contract does not need.\n\n| Question | Required decision |\n|---|---|\n| What does the acknowledgment establish? | Received, durably accepted, executing, committed, or ready; define persistence\u002Ffailure scope |\n| What does the next step require? | Exact resource, generation, desired\u002Fapplied revision, freshness and authorization |\n| What identifies and authorizes work? | Separate request\u002Foperation ID, resource identity, runtime boot\u002Fconnection generation, revision, and authority\u002Ffencing |\n| Who knows the outcome? | Authoritative query and its consistency, retention, unknown\u002Fexpired\u002Fnot-found meanings; event recovery if used |\n| Who may retry or clean up? | One coordinating retry owner, mutation idempotency contract, receiver-side ownership check, cancellation\u002Fcompensation owner |\n| What bounds waiting and execution? | Per-call limit, remaining end-to-end budget, no-progress definition, backend lifetime and cleanup lifetime |\n\n## Choose the minimal sufficient protocol\n\nA bounded synchronous call is enough when it returns the required committed\nresult with documented semantics. An existing authoritative read plus bounded\npolling may be simpler than a watch. Long-running work should acknowledge\nacceptance with a queryable operation handle rather than hold the initial RPC\nuntil readiness. Distinguish volatile receipt from durable acceptance explicitly.\nDo not add streaming, a broker, a database or a general operation framework\nunless the boundary's requirements actually need it.\n\nApply the following invariants to the affected path:\n\n1. **Identity is not authority.** Correlate replies to their request, resource,\n   boot\u002Fconnection generation and revision as applicable. A UUID, PID, open port\n   or matching config is not permission. Preserve credential and provenance\n   checks. The receiver must atomically validate current ownership\u002Ffencing with\n   its side effect, through its existing serialization\u002Fconditional-write seam.\n   A client precheck alone leaves a race.\n2. **State establishes correctness; events accelerate observation.** Use an\n   authoritative query with adequate freshness\u002Fconsistency. Snapshot+watch needs\n   a producer-supported revision\u002Fcursor bridge, or subscribe\u002Fbuffer plus a\n   documented snapshot barrier. Recover by resnapshot on reconnect or expired\n   history; reject duplicate, reordered and stale-generation events. Never\n   assume arbitrary events have etcd's delivery guarantees. Without a safe\n   bridge use bounded polling if the query suffices.\n3. **Operation outcome and runtime health differ.** Terminal operation outcomes\n   are immutable for their documented retention lifetime. Observation sequence\n   increases within its declared scope, but health\u002Freadiness may regress.\n   A new boot is a new generation, even on the same endpoint. An old successful\n   launch does not prove that the current runtime is ready.\n4. **Timeout is an observation result.** Before dispatch, expiry may prevent\n   work. After dispatch, timeout\u002Fdisconnect may leave the outcome unknown,\n   including effects already committed. Keep that uncertainty separate from\n   authoritative terminal failure. Query by identity; distinguish history\n   expiry from proof of absence. Replay only under an applicable idempotency\n   contract. Never roll back merely because an acknowledgment was lost.\n5. **Bound every call and the whole wait.** Propagate the remaining end-to-end\n   duration, accounting for queueing, connection\u002Fauthentication, reads, polling\n   and backoff. Measure local elapsed time monotonically; do not compare\n   unsynchronized process clocks as a deadline protocol. Use existing named,\n   configurable repository limit patterns with measured\u002Fjustified values.\n   Responsiveness heartbeat is not useful progress. Only documented forward\n   progress renews a no-progress allowance; nothing renews the outer ceiling.\n   Client wait expiry does not implicitly terminate accepted backend work.\n   Specify backend execution, result retention and cleanup ownership separately.\n6. **Idempotency must be implemented, not inferred.** Establish caller\u002Fresource\n   scope, same-key\u002Fdifferent-parameters conflict, concurrent deduplication,\n   retention\u002Fexpiry and the crash-safe relationship of effect and result record.\n   If external effects cannot commit atomically with the record, require\n   recoverable intent\u002Freconciliation and name residual ambiguity. UUIDs and\n   in-memory caches do not establish durable exactly-once execution.\n7. **Retry at one coordinating layer.** Inventory SDK, transport and workflow\n   retries to avoid multiplication. Define bounded attempts\u002Frate, transient\n   classification, capped backoff with jitter and supported retry hints within\n   the remaining budget. Query retries and mutation retries are different;\n   authentication\u002Fconflict\u002Fprovenance failures are not generic transient errors.\n8. **Cancellation is not rollback.** Distinguish cancellation requested,\n   execution stopped, descendants drained and cleanup\u002Fcompensation complete.\n   Completion may win the race. Query the actual terminal outcome. Compensation\n   must be explicit, idempotent, resumable and scoped to current authority at\n   the mutation boundary. Never restore an old snapshot over concurrent edits.\n9. **Prove the next step, not a proxy.** Saved config is not applied config.\n   Alive, listener owned, startup ready, operator admitted and capability\n   approved are different predicates. Bind the required predicate to current\n   generation\u002Frevision\u002Ffreshness and the actual caller's authorization.\n   Current matching state can permit a next step when the contract says so;\n   it does not prove which historical operation caused it.\n\n## Inspect producer support and disposition gaps\n\nUse the existing supported contract, including capability\u002Fversion negotiation,\nbefore declaring a feature missing. Implement only consumer behavior justified\nby the supported producer versions. An honest compatibility path may offer a\nweaker explicitly named guarantee if the next step accepts it. Otherwise report\nthe precise unavailable confirmation and stop that transition, not the whole\nunrelated product.\n\nFor each genuine producer gap, record:\n\n- Repository, authoritative producer file\u002Fsymbol, pinned SHA and source\u002Ftest\n  links; distinguish implementation, documented guarantee, runtime proof and\n  installed-version unknowns.\n- Required missing contract, existing support checked, and why consumer reads,\n  local IDs or timing cannot establish it. Separate historical attribution from\n  present-state reconciliation.\n- Minimal versioned producer proposal: schema\u002Fsemantics, authorization\u002Ffencing,\n  persistence\u002Fretention, query\u002Frecovery and errors only as needed for this path.\n- Compatibility and rollout: capability discovery, producer-first deployment,\n  supported older versions, consumer gating and explicit blocked\u002Fdegraded state.\n  Do not guess a minimum released version from a source commit.\n- Acceptance tests covering the failure window that requires producer support.\n  Suggest the update to its owning repository; do not silently implement a\n  consumer workaround or mutate another repository without authority.\n\nThe [pinned onboarding case study](references\u002Fupstream-gaps.md) illustrates this\ndisposition, not a permanent capability catalog. Recheck it for the task's\nactual producer revision. [Research notes](references\u002Fprinciples.md) explain\nthe source-supported principles without importing vendor-specific requirements.\n\nNo sleeps, longer timeouts, log parsing, unconditional cleanup or speculative\nfallback may substitute for required producer guarantees. Diagnostic logs can\nexplain a failure, not become an undocumented protocol. Preserve necessary\ntimeouts and all credential\u002Fprovenance checks.\n\n## Select failure proof and hand off\n\nUse the existing test harness and the [test authoring](..\u002Ftest-audit\u002FSKILL.md),\n[test selection](..\u002Fopenclaw-testing\u002FSKILL.md) and\n[proof](..\u002Fopenclaw-proof-validation\u002FSKILL.md) routes. Select applicable cases,\nincluding receiver-side tests where a consumer mock cannot prove the guarantee:\n\n| Failure injection | Observable assertion |\n|---|---|\n| Drop acknowledgment after acceptance\u002Fcommit | Reconcile same identity; no blind mutation replay or false terminal failure |\n| Duplicate\u002Freorder events; disconnect at snapshot\u002Fsubscribe | No missed required transition or stale state regression; recover from reconnect\u002Fhistory expiry |\n| Crash between effect and receipt; concurrent same key | Recover intent\u002Foutcome without duplicate effect; reject changed parameters; expose expiry ambiguity |\n| Stale cleanup or delayed write after a new owner | Receiver rejects obsolete authority at the side-effect boundary; newer state survives |\n| Cancel as completion commits; child still running | Terminal winner preserved; requested\u002Fstopped\u002Fdrained\u002Fcompensated remain distinct |\n| New boot reuses PID\u002Fport; stale config\u002Fhealth reply | Reject stale generation\u002Frevision; verify fresh own-identity admission as required |\n| Exact timeout\u002Fno-progress boundaries with virtual clocks | Finite calls and outer ceiling; wall-clock jumps\u002Fheartbeats do not extend budget; pre\u002Fpost-dispatch outcomes differ |\n| Retry storm, unsupported producer, authorization failure | One bounded retry owner; no multiplied retries or invented compatibility success |\n\nFor a small read-only query change, retain only the relevant invariants\u002Fcases\nand explain exclusions; do not demand a full onboarding E2E suite. For runtime\nclaims, collect current-head evidence at the real changed boundary or name the\nblocker. A fixture proves only its simulated contract.\n\nHandoff: changed owner and invariant, exact source\u002Fbuild\u002Fversion evidence,\nconfirmation actually achieved, commands\u002Fresults, unverified boundaries and\nupstream dispositions. Do not claim stronger guarantees than the producer.\n",{"name":5,"description":7},"main","a6ff6a164b20835bd5d77200c9134578dd7ce0c6",".agents\u002Fskills\u002Fnetwork-handoff",349,"https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node","https:\u002F\u002Fopenclaw.ai","https:\u002F\u002Fgithub.com\u002Fopenclaw","Your personal, open source AI assistant.","2026-11-14 18:13:49",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add openclaw\u002Fopenclaw-windows-node --skill network-handoff",{"tool":38,"command":39},"bunx","bunx skills add openclaw\u002Fopenclaw-windows-node --skill network-handoff",{"tool":41,"command":42},"pnpm","pnpm dlx skills add openclaw\u002Fopenclaw-windows-node --skill network-handoff","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fnetwork-handoff\u002FSKILL.md",true,[46,52],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fprinciples.md","REFERENCE","参考文档",3839,"https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fnetwork-handoff\u002Freferences\u002Fprinciples.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fupstream-gaps.md",17097,"https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fnetwork-handoff\u002Freferences\u002Fupstream-gaps.md",[57,83,102,121,142,163],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":82},"01M6KZCXK5PCYWWW5FCVDYPAXK","retrigger-flaky-ci","Triage failed CI on a pull request or the active branch, decide whether the failures come from its changes, and rerun only failures that are unrelated. Use when asked to retrigger, rerun, or kick CI on a PR with failing checks.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fretrigger-flaky-ci\u002FSKILL.md",[],[64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81],"retrigger flaky ci","ci","when","unrelated","failing","rerun","use","triage","only","decide","request","active","pull request","pr","changes","branch","with","pull","2026-11-10 10:12:30",{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":82},"01M698QMQN5TQES4R7ZPTYAZF5","technical-documentation","Write, review, or restructure Windows node documentation and agent instructions using current source, task-focused explanations, and explicit fact preservation.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftechnical-documentation\u002FSKILL.md",[],[90,91,92,93,94,95,96,97,98,99,100,101],"technical documentation","technical","documentation","agent instructions","windows","source","preservation","node","write","instructions","review","task",{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":82},"01M698QKPXJ8ZG34MYHPPCAT4T","pr-simplify","Simplify your own Windows-node change to its smallest complete fix while preserving distinct regressions, platform contracts, and required proof. Use after implementation before PR preparation, after large repairs, or when asked to shrink a change or remove slop.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fpr-simplify\u002FSKILL.md",[],[109,77,110,111,112,113,114,115,97,116,66,70,117,118,119,94,120],"pr simplify","simplify","contracts","slop","proof","implementation","preparation","change","fix","complete","remove","platform",{"skillNo":122,"name":123,"nameZh":6,"description":124,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":132,"publishedAt":82},"01M698QJEA384YPM8R9CPZ1YFD","openclaw-testing","Choose proportional OpenClaw Windows node tests, builds, packaging checks, and runtime proof, then diagnose failures without weakening required gates.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fopenclaw-testing\u002FSKILL.md",[127],{"categoryNo":128,"categoryName":129,"taxonomyType":130,"taxonomyTypeDesc":131},"01JBSEEDCAT000000000000008","测试与质量","SCENARIO","使用场景",[133,9,134,94,113,97,135,136,137,138,139,140,141],"testing","openclaw testing","gates","runtime proof","tests","choose","packaging","runtime","diagnose",{"skillNo":143,"name":144,"nameZh":6,"description":145,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":146,"categories":147,"keywords":151,"publishedAt":82},"01M698QH7JVVA09769ER9P2TKY","openclaw-proof-validation","Plan and collect OpenClaw Windows validation\u002Fproof: tests, rubber-duck review, UI evidence, MCP output, and gateway runtime proof.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fopenclaw-proof-validation\u002FSKILL.md",[148],{"categoryNo":149,"categoryName":150,"taxonomyType":130,"taxonomyTypeDesc":131},"01JBSEEDCAT000000000000023","Skill与智能体开发",[152,9,153,113,154,94,155,156,140,157,158,159,160,161,100,136,137,162],"openclaw proof validation","validation","evidence","gateway","MCP","duck","ui","ui evidence","plan","output","collect",{"skillNo":164,"name":165,"nameZh":6,"description":166,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":167,"categories":168,"keywords":174,"publishedAt":82},"01M698QFXW9PDK8YZ467MBW5P1","openclaw-debugging","Debug OpenClaw Windows Companion, setup, Gateway connection, node capabilities, packaged permissions, updates, installers, and system.run by selecting the failing boundary and proof path before changing code.","https:\u002F\u002Fgithub.com\u002Fopenclaw\u002Fopenclaw-windows-node\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fopenclaw-debugging\u002FSKILL.md",[169],{"categoryNo":170,"categoryName":171,"taxonomyType":172,"taxonomyTypeDesc":173},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[175,176,9,94,177,113,178,179,180,181,155,182,183,97,184,185,186,187,188,68],"openclaw debugging","debugging","setup","connection","boundary","debug","system","path","companion","run","permissions","capabilities","code","updates"]