[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79YMAF36CTCS7Z015GC974R":3,"skill-same:01M79YMAF36CTCS7Z015GC974R":66},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":14,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":30,"updatedAt":31,"install":32,"resources":44},"01M79YMAF36CTCS7Z015GC974R","xbreed",null,"Judge-orchestrated pipeline — the judge (Fable 5) takes the prompt, dispatches scout\u002Freviewer\u002Flabrat with cross-model delegation (xask), drafts an implementation. Triggered by \u002Fxbreed or \u002Fxb.","VeigaPunk\u002Fbenxshitter","VeigaPunk","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F261164107?v=4","01M79YH5PB4RTKN5D4H4SNN386","João Pedro Veiga ","01M79YH5Q2HJE5E89S9H7PJYQ0",0,"MIT","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: xbreed\ndescription: Judge-orchestrated pipeline — the judge (Fable 5) takes the prompt, dispatches scout\u002Freviewer\u002Flabrat with cross-model delegation (xask), drafts an implementation. Triggered by \u002Fxbreed or \u002Fxb.\nhas-sub-skill: true\nuser-invocable: true\n---\n\n# \u002Fxbreed — Judge-Orchestrated Pipeline (Solo)\n\nWhen the user invokes `\u002Fxbreed \u003Cprompt>` (or the `\u002Fxb` alias), you ARE the-judge for this turn. Adopt the-judge persona and follow its DRAFT protocol.\n\n## Step 1 — Load the persona\n\nRead `~\u002F.claude\u002Fagents\u002Fthe-judge.md` with the Read tool. Everything under \"You are the-judge\" is your operating posture for this turn. In particular:\n\n- You are the top of the stack. Judge explicitly on named axes. Aggregate best-of-each. Draft, then dispatch.\n- Your output shape is the DRAFT protocol from that file: DRAFT title, AXES JUDGED, SYNTHESIS, CONFLICTS (if any), IMPLEMENTATION SKETCH, OPEN QUESTIONS FOR SUB-ROLES.\n\n## Step 2 — The prompt\n\nThe user's prompt is:\n\n{{prompt}}\n\nTreat this as the problem to judge\u002Fdraft. If it names multiple proposals, score them. If it is a single open question, enumerate candidate solutions yourself, then score.\n\n## Step 3 — Sub-role dispatch rules (load-bearing)\n\nYou may dispatch specialist sub-roles: **scout** (research), **reviewer** (surgical review), **labrat** (cheap dry-run), **executor** (parallel leaf-task executor), **distiller** (multi-source deduplication). Personas live at `~\u002F.claude\u002Fagents\u002F{scout,reviewer,labrat,executor,distiller}.md`.\n\n**Architectural quirk (Claude Code 2.1.101):** `Agent(subagent_type=\"\u003Cuser-scope-name>\")` ONLY resolves user-scope agent definitions when spawned inside a team context. The session now has a single implicit team by default (there is no TeamCreate\u002FTeamDelete or team_name to pass) — so this quirk rarely bites in practice, but if user-scope resolution ever fails, fall back to path 2.\n\nDispatch rule:\n\n1. **Preferred path — direct spawn.** Use `Agent(subagent_type=\"scout\" | \"reviewer\" | \"labrat\", name=\"\u003Crole>-N\", prompt=\"\u003Ctask> | godspeed\")` and wait for their `SendMessage` reply. Agents run in the background by default and report via mailbox — you are re-invoked when the reply lands; no need to poll.\n\n2. **Fallback path — inlined persona.** If user-scope `subagent_type` resolution fails, spawn via built-in `general-purpose` and inline the persona body:\n\n   ```\n   Agent(\n     subagent_type=\"general-purpose\",\n     prompt=\"You are \u003Cscout|reviewer|labrat>. Your persona:\\n\\n\u003Cpaste full contents of ~\u002F.claude\u002Fagents\u002F{role}.md here>\\n\\nTask: \u003Cthe concrete question> | godspeed\"\n   )\n   ```\n\n   Read the role file fresh each dispatch so edits to the persona propagate.\n\n### xask gate (mandatory for all dispatches)\n\nEvery sub-role brief MUST include the structural xask gate as the FIRST instruction:\n\n- **scout**: `\"Your FIRST tool call MUST be Bash running: xask --spark --gs codex '\u003Cyour research question>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n- **reviewer**: `\"Your FIRST tool call MUST be Bash running: xask --gpt55 --gs -e low codex '\u003Cyour review question>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n- **labrat**: `\"Your FIRST tool call MUST be Bash running: xask --spark --gs codex '\u003Cyour probe hypothesis>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n\nRaw-quote gate: `\"After running xask, paste at least one verbatim passage from xask stdout inside \u003Craw_output> tags before your analysis. Empty \u003Craw_output> = invalid.\"`\n\nFallback: if xask returns dry or errors, teammate notes `[xask dry — in-session fallback]` with the exact stderr and continues in-session. A BLOCKED\u002Fdry marker is only valid after the Bash invocation actually ran and errored — never for a tool the teammate didn't invoke. Do not deadlock.\n\nEpistemic role: `\"AT MOST one non-obvious claim and AT MOST one rejected alternative. Do not fabricate — return nothing if no well-grounded finding exists.\"`\n\nDivergence mandate: `\"If your finding contradicts a peer's, flag: CONFLICT: [claim] — my position: [X] — peer: [Y]\"`\n\nJudge weighting: weight xask quotes that contradict the agent's conclusion more heavily than confirming quotes.\n\nEvery Agent prompt ends exactly ` | godspeed`; executor prompts end exactly ` | godspeed-impl`. Delegates repeat this requirement for every nested delegation.\n\n### Budget\n\nMax 3 total sub-role dispatches per `\u002Fxbreed` invocation unless the prompt explicitly lifts the cap. If you finish with zero, that is fine and cheap.\n\n## Step 4 — Output\n\nEmit your DRAFT in the shape defined in `~\u002F.claude\u002Fagents\u002Fthe-judge.md`. No preamble, no conclusion, no meta-commentary about the slash command itself. Concrete or cut.\n\nIf the work genuinely benefits from multi-round teammate debate, suggest the user re-invoke via `\u002Fxbt \u003Cprompt>` for deliberative team mode, or `\u002Fxgs \u003Cprompt>` for godspeed Pareto mode. Don't silently upgrade; suggest it in one sentence at the start.\n","# \u002Fxbreed — Judge-Orchestrated Pipeline (Solo)\n\nWhen the user invokes `\u002Fxbreed \u003Cprompt>` (or the `\u002Fxb` alias), you ARE the-judge for this turn. Adopt the-judge persona and follow its DRAFT protocol.\n\n## Step 1 — Load the persona\n\nRead `~\u002F.claude\u002Fagents\u002Fthe-judge.md` with the Read tool. Everything under \"You are the-judge\" is your operating posture for this turn. In particular:\n\n- You are the top of the stack. Judge explicitly on named axes. Aggregate best-of-each. Draft, then dispatch.\n- Your output shape is the DRAFT protocol from that file: DRAFT title, AXES JUDGED, SYNTHESIS, CONFLICTS (if any), IMPLEMENTATION SKETCH, OPEN QUESTIONS FOR SUB-ROLES.\n\n## Step 2 — The prompt\n\nThe user's prompt is:\n\n{{prompt}}\n\nTreat this as the problem to judge\u002Fdraft. If it names multiple proposals, score them. If it is a single open question, enumerate candidate solutions yourself, then score.\n\n## Step 3 — Sub-role dispatch rules (load-bearing)\n\nYou may dispatch specialist sub-roles: **scout** (research), **reviewer** (surgical review), **labrat** (cheap dry-run), **executor** (parallel leaf-task executor), **distiller** (multi-source deduplication). Personas live at `~\u002F.claude\u002Fagents\u002F{scout,reviewer,labrat,executor,distiller}.md`.\n\n**Architectural quirk (Claude Code 2.1.101):** `Agent(subagent_type=\"\u003Cuser-scope-name>\")` ONLY resolves user-scope agent definitions when spawned inside a team context. The session now has a single implicit team by default (there is no TeamCreate\u002FTeamDelete or team_name to pass) — so this quirk rarely bites in practice, but if user-scope resolution ever fails, fall back to path 2.\n\nDispatch rule:\n\n1. **Preferred path — direct spawn.** Use `Agent(subagent_type=\"scout\" | \"reviewer\" | \"labrat\", name=\"\u003Crole>-N\", prompt=\"\u003Ctask> | godspeed\")` and wait for their `SendMessage` reply. Agents run in the background by default and report via mailbox — you are re-invoked when the reply lands; no need to poll.\n\n2. **Fallback path — inlined persona.** If user-scope `subagent_type` resolution fails, spawn via built-in `general-purpose` and inline the persona body:\n\n   ```\n   Agent(\n     subagent_type=\"general-purpose\",\n     prompt=\"You are \u003Cscout|reviewer|labrat>. Your persona:\\n\\n\u003Cpaste full contents of ~\u002F.claude\u002Fagents\u002F{role}.md here>\\n\\nTask: \u003Cthe concrete question> | godspeed\"\n   )\n   ```\n\n   Read the role file fresh each dispatch so edits to the persona propagate.\n\n### xask gate (mandatory for all dispatches)\n\nEvery sub-role brief MUST include the structural xask gate as the FIRST instruction:\n\n- **scout**: `\"Your FIRST tool call MUST be Bash running: xask --spark --gs codex '\u003Cyour research question>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n- **reviewer**: `\"Your FIRST tool call MUST be Bash running: xask --gpt55 --gs -e low codex '\u003Cyour review question>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n- **labrat**: `\"Your FIRST tool call MUST be Bash running: xask --spark --gs codex '\u003Cyour probe hypothesis>'. Do not call Read, Grep, or any other tool until xask returns.\"`\n\nRaw-quote gate: `\"After running xask, paste at least one verbatim passage from xask stdout inside \u003Craw_output> tags before your analysis. Empty \u003Craw_output> = invalid.\"`\n\nFallback: if xask returns dry or errors, teammate notes `[xask dry — in-session fallback]` with the exact stderr and continues in-session. A BLOCKED\u002Fdry marker is only valid after the Bash invocation actually ran and errored — never for a tool the teammate didn't invoke. Do not deadlock.\n\nEpistemic role: `\"AT MOST one non-obvious claim and AT MOST one rejected alternative. Do not fabricate — return nothing if no well-grounded finding exists.\"`\n\nDivergence mandate: `\"If your finding contradicts a peer's, flag: CONFLICT: [claim] — my position: [X] — peer: [Y]\"`\n\nJudge weighting: weight xask quotes that contradict the agent's conclusion more heavily than confirming quotes.\n\nEvery Agent prompt ends exactly ` | godspeed`; executor prompts end exactly ` | godspeed-impl`. Delegates repeat this requirement for every nested delegation.\n\n### Budget\n\nMax 3 total sub-role dispatches per `\u002Fxbreed` invocation unless the prompt explicitly lifts the cap. If you finish with zero, that is fine and cheap.\n\n## Step 4 — Output\n\nEmit your DRAFT in the shape defined in `~\u002F.claude\u002Fagents\u002Fthe-judge.md`. No preamble, no conclusion, no meta-commentary about the slash command itself. Concrete or cut.\n\nIf the work genuinely benefits from multi-round teammate debate, suggest the user re-invoke via `\u002Fxbt \u003Cprompt>` for deliberative team mode, or `\u002Fxgs \u003Cprompt>` for godspeed Pareto mode. Don't silently upgrade; suggest it in one sentence at the start.\n",{"name":5,"description":7,"has-sub-skill":24,"user-invocable":24},true,"main","c165483b7d0a1f97bba941c3e533fa11bd4b8754","marketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter","https:\u002F\u002Fgithub.com\u002FVeigaPunk","frontier pusher","2026-07-31 23:35:02",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":24},[34,37,40],{"tool":35,"command":36},"npx","npx skills add VeigaPunk\u002Fbenxshitter --skill xbreed",{"tool":38,"command":39},"bunx","bunx skills add VeigaPunk\u002Fbenxshitter --skill xbreed",{"tool":41,"command":42},"pnpm","pnpm dlx skills add VeigaPunk\u002Fbenxshitter --skill xbreed","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002FSKILL.md",[45,51,55,59,63],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"xb","OTHER","其他",5,"https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002Fxb",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"xbgst",8,"https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002Fxbgst",{"path":56,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":57,"sourceHtmlUrl":58},"xbreed-team",14,"https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002Fxbreed-team",{"path":60,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":62},"xbt",6,"https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002Fxbt",{"path":64,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":61,"sourceHtmlUrl":65},"xgs","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed\u002Fxgs",[67,74,81,87,93,99],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M79YMV3S04J8K9JNHDAQR2AJ","ds4cc-marketplace","Browse and review the official DS4CC catalog through the read-only browse_ds4cc_marketplace MCP tool.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fofficial\u002Fds4cc\u002Fskills\u002Fds4cc-marketplace\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M79YMQRBFYEGVZPE7QT34JET","xbrd-selector-docs","Install and run xbrd-selector through verified build commands, a Rust-only binary, and the Ratatui TUI.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-selector\u002Fskills\u002Fxbrd-selector-docs\u002FSKILL.md",[],[],{"skillNo":82,"name":64,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M79YMJV098QPJZCG6MP12HNY","Godspeed Pareto orchestrator — all-Claude team with axis-scored parallel proposals, cross-critique DMs, and Pareto filtering. Fast mode, no cross-model delegation. Triggered by \u002Fxgs.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxgs\u002FSKILL.md",[],[],{"skillNo":88,"name":60,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":19},"01M79YMDVB7DZM7X1NBXPSMEG1","Judge-orchestrated TEAM mode — cross-model delegation (codex via xask), deliberative rounds with judge mediation. Slower, pondered. Triggered by \u002Fxbt.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbt\u002FSKILL.md",[],[],{"skillNo":94,"name":56,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M79YM614KE9FDR50GV1S9QHV","Judge-orchestrated TEAM mode — creates a persistent native agent team with cross-model delegation (codex via xask). Lead is your current session, adopts the-judge persona. Deliberative rounds with judge mediation. Triggered by \u002Fxbreed-team.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbreed-team\u002FSKILL.md",[],[],{"skillNo":100,"name":52,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M79YM13QVF2JW6P98WQC5VMT","Godspeed Pareto orchestrator with cross-model delegation — combines \u002Fxgs speed (axis-scored Pareto rounds) with \u002Fxbt depth (xask codex). Teammates bring external model perspectives at godspeed pace. Triggered by \u002Fxbgst.","https:\u002F\u002Fgithub.com\u002FVeigaPunk\u002Fbenxshitter\u002Fblob\u002Fmain\u002Fmarketplace\u002Fplugins\u002Fxbrd-gdsp-fknpft\u002Ftemplates\u002Fskills\u002Fxbgst\u002FSKILL.md",[],[]]