[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75X7CVQDM2HJ83G83CJF1CD":3,"skill-same:01M75X7CVQDM2HJ83G83CJF1CD":44},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"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":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M75X7CVQDM2HJ83G83CJF1CD","pk-subagent-orchestrator",null,"kingkillery\u002Foh-my-pi","kingkillery","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F200727508?v=4","01M6X59E9Q2FCEGBVQ1P6FT0YC","PK","01M75X46A278QMZ5PC4PH2QYJG",2,"MIT","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fverifier-extension\u002Fskills\u002Fpk-subagent-orchestrator\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: pk-subagent-orchestrator\ndescription: Use when designing, configuring, or running Oh My Pi subagent orchestration: one visible endpoint over task\u002Fquick_task\u002Foracle\u002Freviewer\u002Fexplore agents, deterministic route planning, parallel waves, recursive delegation caps, independent verification, and final synthesis without exposing internal routing. Also use when invoking the verifier-extension `subagent_orchestrator_plan` tool to compute an OMP-native route plan.\n---\n\n# PK Subagent Orchestrator for Oh My Pi\n\nBuild one public controller over Oh My Pi subagents. The caller sees one answer. Internally you MAY answer directly, call one specialist, spawn a parallel `task` wave, recurse through a narrower child orchestrator, verify, then synthesize.\n\n\u003Ccritical>\n- NEVER delegate simple low-risk requests.\n- NEVER expose routing, subagent names, recursion depth, or raw transcripts.\n- NEVER treat subagent output as truth. Evidence wins.\n- MUST cap recursion and fanout before spawning.\n- MUST synthesize one answer in the user's requested format.\n\u003C\u002Fcritical>\n\n## OMP primitives\n\n| Need | OMP primitive |\n|---|---|\n| Specialist implementation\u002Fresearch | `task` |\n| Mechanical lookup\u002Fupdate | `quick_task` |\n| Broad read-only codebase scouting | `explore` |\n| Senior hands-on judgment | `oracle` |\n| Review\u002Ffailure finding | `reviewer` or CE reviewer agents |\n| Parallel wave | one `task` call with `tasks[]` |\n| Peer coordination | IRC + Agent Hub |\n| Route-plan computation | verifier-extension `subagent_orchestrator_plan` |\n| Independent verification | `llm_as_verifier`, `fmh`, reviewer from different model family |\n\n## Route first\n\nBefore spawning, classify:\n\n- complexity: `single-step | multi-step | open-ended`\n- risk: `low | med | high`\n- evidence need: `current-context | tool-retrieval | multi-source`\n- decomposition: `independent | sequential | not-decomposable`\n- data sensitivity: `public | internal | confidential | unknown`\n\nIf the extension tool is available, call `subagent_orchestrator_plan` with:\n\n```json\n{\n  \"request\": \"\u003Cuser request or narrowed subproblem>\",\n  \"complexity\": \"multi-step\",\n  \"risk\": \"med\",\n  \"evidenceNeed\": \"tool-retrieval\",\n  \"decomposability\": \"independent\",\n  \"dataSensitivity\": \"unknown\",\n  \"recursiveAllowed\": false,\n  \"specialists\": [\n    {\n      \"name\": \"CodeResearcher\",\n      \"scope\": \"codebase discovery and file evidence\",\n      \"costTier\": \"low\",\n      \"role\": \"specialist\",\n      \"capabilities\": [\"read\", \"search\", \"codegraph\"]\n    },\n    {\n      \"name\": \"IndependentVerifier\",\n      \"scope\": \"evidence checking and contradiction finding\",\n      \"costTier\": \"med\",\n      \"role\": \"verifier\"\n    }\n  ]\n}\n```\n\nTool result gives:\n\n- `mode`: `fast | deep`\n- `routing`: `direct | single | parallel | recursive`\n- `subagents`: selected specialists\n- `verification`: `V0..V3`\n- `maxDepth`, `maxFanout`, `childCallLimit`\n- private hidden route-plan block\n\nUse the plan. Override only when repo evidence proves it wrong.\n\n## Routing policy\n\n- Direct: simple + low risk + current context sufficient.\n- Single: one specialist materially improves result.\n- Parallel: independent subtasks can run without shared output.\n- Sequential: B requires A's artifact or decision.\n- Recursive: only open-ended decomposable subproblem needs orchestration.\n\nCost rule: cheapest capable first. Escalate on high risk, weak evidence, conflict, verifier failure, missing capability, or human-only decision.\n\n## Delegation contract\n\nEvery subagent assignment MUST include:\n\n- Objective: one sentence.\n- Scope: exact files\u002Fsystems; explicit non-goals.\n- Inputs: paths, handles, artifacts, constraints.\n- Output: required shape.\n- Evidence: files, tests, sources, or `unknown`.\n- Failure: what to report instead of guessing.\n- Verification ban: subagents do not run project-wide gates.\n\nUse OMP task format:\n\n```markdown\n# Goal\nOne-sentence batch goal.\n\n# Constraints\nMUST\u002FNEVER rules, shared assumptions, no project-wide gates.\n\n# Contract\nShared interfaces, expected outputs, handoff format.\n```\n\nPer task:\n\n```markdown\n# Target\nExact files\u002Fsymbols\u002Fscope.\n\n# Change\nConcrete work or investigation.\n\n# Acceptance\nObservable result; no project-wide commands.\n```\n\n## Verification tiers\n\n- V0 self-check: low-risk subjective output.\n- V1 evidence check: factual\u002Ftechnical\u002Fuser-actionable.\n- V2 independent review: complex, multi-agent, code changes, conflicts.\n- V3 adversarial: security, privacy, legal, financial, destructive, credentials.\n\nV2+ MUST use an independent family\u002Fpersona. Aggregate by evidence, not votes.\n\n## Recursion controls\n\nDefaults:\n\n- max depth: `2`\n- max fanout per level: `5`\n- max total child calls: `12`\n\nStop when solved, uncertainty repeats, two iterations add no evidence, outputs become circular, or more calls will not change the answer.\n\nChild orchestrators receive a narrower objective, allowed subagents, non-goals, output contract, and stop condition. Parent owns final synthesis.\n\n## Final synthesis\n\nReturn one answer:\n\n- Satisfy the user's requested format.\n- Cite files\u002Ftests\u002Fsources when trust depends on them.\n- Mark inference explicitly.\n- Name unresolved blockers.\n- Hide route details and raw transcripts.\n- Do not average contradictions; resolve by evidence.\n\n\u003Ccritical>\n- Route before spawning.\n- Parallelize independent subtasks in one `task` batch.\n- Verify before final synthesis.\n- Keep routing private.\n\u003C\u002Fcritical>\n","# PK Subagent Orchestrator for Oh My Pi\n\nBuild one public controller over Oh My Pi subagents. The caller sees one answer. Internally you MAY answer directly, call one specialist, spawn a parallel `task` wave, recurse through a narrower child orchestrator, verify, then synthesize.\n\n\u003Ccritical>\n- NEVER delegate simple low-risk requests.\n- NEVER expose routing, subagent names, recursion depth, or raw transcripts.\n- NEVER treat subagent output as truth. Evidence wins.\n- MUST cap recursion and fanout before spawning.\n- MUST synthesize one answer in the user's requested format.\n\u003C\u002Fcritical>\n\n## OMP primitives\n\n| Need | OMP primitive |\n|---|---|\n| Specialist implementation\u002Fresearch | `task` |\n| Mechanical lookup\u002Fupdate | `quick_task` |\n| Broad read-only codebase scouting | `explore` |\n| Senior hands-on judgment | `oracle` |\n| Review\u002Ffailure finding | `reviewer` or CE reviewer agents |\n| Parallel wave | one `task` call with `tasks[]` |\n| Peer coordination | IRC + Agent Hub |\n| Route-plan computation | verifier-extension `subagent_orchestrator_plan` |\n| Independent verification | `llm_as_verifier`, `fmh`, reviewer from different model family |\n\n## Route first\n\nBefore spawning, classify:\n\n- complexity: `single-step | multi-step | open-ended`\n- risk: `low | med | high`\n- evidence need: `current-context | tool-retrieval | multi-source`\n- decomposition: `independent | sequential | not-decomposable`\n- data sensitivity: `public | internal | confidential | unknown`\n\nIf the extension tool is available, call `subagent_orchestrator_plan` with:\n\n```json\n{\n  \"request\": \"\u003Cuser request or narrowed subproblem>\",\n  \"complexity\": \"multi-step\",\n  \"risk\": \"med\",\n  \"evidenceNeed\": \"tool-retrieval\",\n  \"decomposability\": \"independent\",\n  \"dataSensitivity\": \"unknown\",\n  \"recursiveAllowed\": false,\n  \"specialists\": [\n    {\n      \"name\": \"CodeResearcher\",\n      \"scope\": \"codebase discovery and file evidence\",\n      \"costTier\": \"low\",\n      \"role\": \"specialist\",\n      \"capabilities\": [\"read\", \"search\", \"codegraph\"]\n    },\n    {\n      \"name\": \"IndependentVerifier\",\n      \"scope\": \"evidence checking and contradiction finding\",\n      \"costTier\": \"med\",\n      \"role\": \"verifier\"\n    }\n  ]\n}\n```\n\nTool result gives:\n\n- `mode`: `fast | deep`\n- `routing`: `direct | single | parallel | recursive`\n- `subagents`: selected specialists\n- `verification`: `V0..V3`\n- `maxDepth`, `maxFanout`, `childCallLimit`\n- private hidden route-plan block\n\nUse the plan. Override only when repo evidence proves it wrong.\n\n## Routing policy\n\n- Direct: simple + low risk + current context sufficient.\n- Single: one specialist materially improves result.\n- Parallel: independent subtasks can run without shared output.\n- Sequential: B requires A's artifact or decision.\n- Recursive: only open-ended decomposable subproblem needs orchestration.\n\nCost rule: cheapest capable first. Escalate on high risk, weak evidence, conflict, verifier failure, missing capability, or human-only decision.\n\n## Delegation contract\n\nEvery subagent assignment MUST include:\n\n- Objective: one sentence.\n- Scope: exact files\u002Fsystems; explicit non-goals.\n- Inputs: paths, handles, artifacts, constraints.\n- Output: required shape.\n- Evidence: files, tests, sources, or `unknown`.\n- Failure: what to report instead of guessing.\n- Verification ban: subagents do not run project-wide gates.\n\nUse OMP task format:\n\n```markdown\n# Goal\nOne-sentence batch goal.\n\n# Constraints\nMUST\u002FNEVER rules, shared assumptions, no project-wide gates.\n\n# Contract\nShared interfaces, expected outputs, handoff format.\n```\n\nPer task:\n\n```markdown\n# Target\nExact files\u002Fsymbols\u002Fscope.\n\n# Change\nConcrete work or investigation.\n\n# Acceptance\nObservable result; no project-wide commands.\n```\n\n## Verification tiers\n\n- V0 self-check: low-risk subjective output.\n- V1 evidence check: factual\u002Ftechnical\u002Fuser-actionable.\n- V2 independent review: complex, multi-agent, code changes, conflicts.\n- V3 adversarial: security, privacy, legal, financial, destructive, credentials.\n\nV2+ MUST use an independent family\u002Fpersona. Aggregate by evidence, not votes.\n\n## Recursion controls\n\nDefaults:\n\n- max depth: `2`\n- max fanout per level: `5`\n- max total child calls: `12`\n\nStop when solved, uncertainty repeats, two iterations add no evidence, outputs become circular, or more calls will not change the answer.\n\nChild orchestrators receive a narrower objective, allowed subagents, non-goals, output contract, and stop condition. Parent owns final synthesis.\n\n## Final synthesis\n\nReturn one answer:\n\n- Satisfy the user's requested format.\n- Cite files\u002Ftests\u002Fsources when trust depends on them.\n- Mark inference explicitly.\n- Name unresolved blockers.\n- Hide route details and raw transcripts.\n- Do not average contradictions; resolve by evidence.\n\n\u003Ccritical>\n- Route before spawning.\n- Parallelize independent subtasks in one `task` batch.\n- Verify before final synthesis.\n- Keep routing private.\n\u003C\u002Fcritical>\n",{},"main","403af0736d24ad919fe5c817fc55f104e6030a8a","packages\u002Fverifier-extension\u002Fskills\u002Fpk-subagent-orchestrator",0,"https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi","https:\u002F\u002Fgithub.com\u002Fkingkillery","2026-07-30 01:42:39",{"source":7,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add kingkillery\u002Foh-my-pi --skill pk-subagent-orchestrator",{"tool":36,"command":37},"bunx","bunx skills add kingkillery\u002Foh-my-pi --skill pk-subagent-orchestrator",{"tool":39,"command":40},"pnpm","pnpm dlx skills add kingkillery\u002Foh-my-pi --skill pk-subagent-orchestrator","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fverifier-extension\u002Fskills\u002Fpk-subagent-orchestrator\u002FSKILL.md",true,[],[45,52,59,64,71,78],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M75X7EWZJ876BEFNVJHT1JM0","llm-as-verifier","This skill should be used when the user asks to \"compare candidate patches\", \"rank multiple solutions\", \"choose between implementations\", \"verify two drafts\", \"pick the best trajectory\", \"use llm-as-a-verifier\", or wants repeated criteria-decomposed LLM verification instead of a single free-form judgment.","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpython\u002Ffugu\u002F.agents\u002Fskills\u002Fllm-as-verifier\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M75X78TNQRB1D04WJMBPYPW6","pk-dynamic-workflows","This skill should be used when the user wants to spawn subagents with different models, configure heterogeneous candidate backends, implement dynamic model routing, or run multi-model fusion workflows. Covers mock, local, premium API (Anthropic\u002FOpenAI), budget API (Kimi\u002FMiniMax\u002FQwen), routing layers (9Router), and CLI backends.","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fverifier-extension\u002Fskills\u002Fpk-dynamic-workflows\u002FSKILL.md",[],[],{"skillNo":60,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":18},"01M75X75C4WSJBJ1YZH0GFJ8E0","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fverifier-extension\u002Fskills\u002Fllm-as-verifier\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":18},"01M75X71TA13A2MQ31KSKPBZ3M","fmh","Run fmh (fusion-meta-harness) — verifier fusion, evaluate-verifier, improve-verifier, promote, run-task. Orchestrates multi-agent candidate generation, rubric scoring, synthesis, and reliability gates for the pi-llm-as-verifier project.","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fverifier-extension\u002Fskills\u002Ffmh\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M75X70G5YJSGJGBYFVDC4ZMY","greet","A greeting skill for testing","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fcoding-agent\u002Ftest\u002Fmarketplace\u002Ffixtures\u002Fvalid-marketplace\u002Fplugins\u002Fhello-plugin\u002Fskills\u002Fgreet\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M75X6ZHTKQNG38AKP18TNMCW","valid-skill","A valid skill for testing purposes.","https:\u002F\u002Fgithub.com\u002Fkingkillery\u002Foh-my-pi\u002Fblob\u002Fmain\u002Fpackages\u002Fcoding-agent\u002Ftest\u002Ffixtures\u002Fskills\u002Fvalid-skill\u002FSKILL.md",[],[]]