@andyrenxu7255

provider-probe audits openai-compatible endpoints to distinguish genuine single-model services from multi-model routing pools, using layered evidence from /models inspection, endpoint compatibility checks, and stability probes for routing decisions.

ars-deep-research orchestrates 13 agents across 6 phases (scoping, investigation, analysis, composition, review, revision) to produce academic research reports. supports 7 modes including socratic dialogue and systematic review. delegates via hermes agent task framework.

12-agent academic paper pipeline supporting 10 modes (full draft, outline, revision, abstract, lit-review, format conversion, citation check, disclosure). handles 6 paper types, 5 citation formats, bilingual abstracts, and multiple output formats via delegated task execution.

poc-battle-card structures competitive analysis during vendor evaluations, producing differentiation matrices, counter-strategies, and talking points. covers poc status tracking, head-to-head comparison grids, and next-step guidance without hallucinating competitor claims.

industry-scenario-generator produces sales collateral (pain points, use cases, ppt materials, talking points) for a given industry and product direction. triggers on scenario/proposal/pitch requests. relies on pre-built knowledge docs but procedure for unmapped industries is underspecified.

office-automation-toolkit is a registry of zero-config cli and python tools for document/pdf/browser automation. it catalogs 20+ libraries with install commands and capability matrices, organized by domain with a quick-reference lookup table.

sop-extractor guides users through structured dialogue to convert daily work procedures into reusable ai skills. supports document parsing and iterative questioning across an 8-step process with zero-installation extraction phase.

meddic-b2b-sales-review organizes b2b pipeline coaching around ptc's six-step framework with gate validation and meddic calibration. it routes scenarios to specialized review templates and references but lacks concrete procedural steps.

defold covers lua-based game development with message passing, dynamic object spawning, gui tooling via druid, physics simulation, sprite animation, input binding, and save/load mechanics. references documentation exists but skill lacks procedural depth.

customer-research automates background investigation and competitor intelligence gathering by executing 12-15 web searches across four dimensions, then generates word reports and before/after scenario ppts. routing logic selects modules based on user intent, but execution clarity and error handling remain underdeveloped.

Forensically verify what model family or routing layer may actually sit behind a claimed LLM endpoint or model ID. Use when an agent must investigate whether...

10-stage end-to-end research-to-publication pipeline orchestrator on Hermes Agent. Stages: planning → deep-research → academic-paper → academic-paper-reviewe...

图片批量加水印。输入图片文件+水印文字，自动在图片上添加水印并返回处理后的图片。 触发词：加水印、图片水印、批量水印。

投标文件合规检查。输入招标文件+投标文件成稿，自动从招标文件提取全部要求并逐项核对投标文件， 标注废标项/资格项/评分项/封装项状态，输出风险汇总和行动清单。 触发词：投标检查、标书核对、标书审查、资格核对、废标检查、投标文件合规。

面向对象的客户记忆体系。聊起任何客户、联系人、项目机会、项目、合同、交付相关话题时激活。 自动维护 memory/customers/ 下的客户索引和详情文件，支持超大客户文件自动分层拆分。 与 sales-visit（明道云写入）双向联动：sales-visit 写入后同步更新本记忆文件， 本记忆文件加载的上下...

投标全流程管理。输入招标文件，自动输出标书CheckList+投标总体计划+投标文件目录结构+单一来源说明（如需）。 触发词：投标、招标、标书、投标管理、投标准备、单一来源。

客户拜访一站式准备。输入客户名+背景+拜访目的，自动输出客户调研报告+拜访策略+话术+注意事项。 触发词：拜访准备、见客户、拜访建议、拜访策略、约访准备。

Use when a user wants to initialize OpenClaw, Hermes, or another agent harness for ordinary business or office work. Guides plain-language business interview...

AI歌曲创作核心技能 — 合辙押韵歌词创作 + MiniMax mmx-cli 正确调用 + 歌词验证 + 歌曲生成全流程。支持古诗词谱曲、主题歌曲创作、多渠道文件投递。

7-agent paper review system on Hermes Agent. 6 modes (full/re-review/quick/methodology-focus/guided/calibration). 5-panel review with editorial decision, rev...