技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
unleash-the-agents【数据与研究】Launch all available agents in parallel waves for open-ended hypothesis generation on problems where the correct domain is unknown. Use when facing a cross-domain problem with no clear starting point, when single-agent approaches have stalled, or when diverse perspectives are more valuable than deep expertise. Produces a ranked hypothesis set with co…
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
並波諮諸客以生多假於開疾——客各以己域為鏡:kabalist 以數、武者以條件分支、靜者以坐覓構。多獨視之合,假之有徵也。
書任客可解之綱(無論域):
## Brief: [Problem Title]
**Problem**: [1-2 sentence statement]
**Examples**:
1. [Input] → [Output] (explain what's known)
2. [Input] → [Output]
3. [Input] → [Output]
4. [Input] → [Output]
5. [Input] → [Output]
**Already tried**: [List failed approaches to avoid rediscovery]
**Success looks like**: [Testable criterion]
**Respond with**:
- Hypothesis: [Your proposed mechanism in one sentence]
- Reasoning: [Why your domain expertise suggests this]
- Confidence: [low/medium/high]
- Testable prediction: [If my hypothesis is correct, then X should be true]
得:綱自足——客僅得此文亦能推。
敗:不能述 5 例或驗法→問未備諮多客。先縮範。
列諸客分為 ~10 一波。前 2 波次序不重;後波之波間注知改善果。
# List all agents from registry
grep ' - id: ' agents/_registry.yml | sed 's/.*- id: //' | shuf
派客於波。初謀 4 波——或不需盡(見早止於四)。
| 波 | 客 | 綱變 |
|---|---|---|
| 1-2 | 20 客 | 標綱 |
| 3 | 10 客 + advocatus-diaboli | 綱+共識+敵挑 |
| 4+ | 各 10 客 | 綱+「X 已確。專邊與敗。」 |
得:派表盡客。advocatus-diaboli 入波 3(非後)以使敵察影響後波。
敗:客不足 20→減 2-3 波。10 客亦行,惟合徵弱。
各波並放。用 sonnet 模以省(值在視多,非個深)。
常互會之路。波中每客以 Agent 工 生為子客(subagent_type 設客類,如 kabalist、geometrist),附綱與域框:
Use the [agent-name] agent to analyze this problem through your domain expertise.
[Paste the brief]
Think about this from your specific perspective as a [agent-description].
[For non-technical agents: add a domain-specific framing, e.g., "What patterns
does your tradition recognize in systems that exhibit this kind of threshold behavior?"]
Respond exactly in the requested format.
並放波中諸客,以 Agent 附 run_in_background: true,以 SendMessage 協於會之單隱團。波畢乃放次波(以波間注知於步四)。
TeamCreate 已棄、閘於常互會外——彼處 ToolSearch("select:TeamCreate") 無返,team_name 被略(會唯一隱團)。其現處(FleetView/雲)增名團物附員任屬與關命(Task* 工於互會亦行——獨團域不行):
TeamCreate({ description: "Wave 1: open-ended hypothesis generation" })TaskCreate(綱+域框)Agent 工生為團員,subagent_type 設客類TaskUpdate 附 owner;察以 TaskListSendMessage({ type: "shutdown_request" })、啟次團附更綱(步四)內建團協——員屬與命、態續於波——然團域獨存於 TeamCreate 現處。互會用法甲。
| Agent 工(法甲) | TeamCreate(法乙) | |
|---|---|---|
| 可得 | 諸會(主) | 獨 FleetView/雲(閘) |
| 宜 | 諸互會之放 | 雲行欲共任跡 |
| 協 | SendMessage、手集 | 任跡、訊、屬 |
| 波接 | 以更綱跡 | 任態續 |
得:每波 2-5 分內返 ~10 結構應。失應或誤式者錄而不阻流。
敗:>50% 波失→察綱清。常因:出式含混、或例不足以使外域客推。
波 1-2 後,未放下波先取現信。
**Update from prior waves**: [N] agents independently proposed [hypothesis family].
Build on this — what explains the remaining cases where this hypothesis fails?
Do NOT simply restate this finding. Extend, challenge, or refine it.
早止建:非每放需盡客。明域問(如庫析)合穩於 30-40 客。抽象開問(如未知數變)全集有值因正域真難測。每波後察合——頂族計與空模比平→後波回減。
此免重發(後波獨重得前果)、引後客向問之邊。
得:後波生更精、更針之假,補現共識之隙。
敗:2 波後無合→問或過開。考縮範或增例。
諸波畢,集應於一文。按機分族去重:
得:排假族表附合計、貢客、代驗預。
敗:每假獨(無合)→信噪過低。或問須更例、或客須更緊出式。
試頂假於空模以確合有意,非訓共產之偽。
得:頂假族顯超偶合且/或過程驗。
敗:頂假敗驗→察次族。皆無過→問或須異法(深單專析、更數、改例)。
宜時:波 3,非後合。 納 advocatus-diaboli 於波 3(與波間知並)優於後波獨敵。早挑使波 4+ 對挑精煉,非堆於未挑共識。
若敵已於波 3,此步為末察。否(如諸波無之)→今生 advocatus-diaboli(或 senior-researcher)。為結構行,以 Agent 工並生二客為子客、以 SendMessage 協之對共識:
Here is the consensus hypothesis from [N] independent agents:
[Hypothesis]
[Supporting evidence and convergence stats]
Your job: find the strongest counterarguments. Where does this fail?
What alternative explanations are equally consistent with the evidence?
What experiment would definitively falsify this hypothesis?
得:反論集、邊例、否驗。假過敵察→可整。佳敵察有時部分護共識——覓設優於替而未善。
敗:敵覓致命瑕→注挑於針隨波(三層+迭式——擇 5-10 客最宜應特挑)。
放覓問;團解之。化驗假族為可行議、組焦團解之。
create-github-issues)subagent_type)生其員為子客、以 SendMessage 協:
teams/ 中合域團定→用之opaque-team(N shapeshifters 適派)——應未知問形不需自定組advocatus-diaboli、contemplative)——彼覺技客漏之施險得:每假族對跡議附團派。放生診;團出修。
敗:團組不合問→重派。Shapeshifter 客可研設而無書工——團頭須施其碼建。
docs/investigations/lead-support-coordination-2026-09-15.mdforage-solutions — 蟻集優以探解空(補:窄範深探)build-coherence — 蜂民主擇諸法(用後此技以擇頂假)coordinate-reasoning — 痕協以管客間訊流coordinate-swarm — 廣集協式於分系expand-awareness — 開覺前縮(補:用為個客備)meditate — 放前清境噪(建於步一前)