vimalinx

Use when joint-genotyping one or more germline gVCFs into a cohort VCF with GATK GenotypeGVCFs.

Use when running GATK HaplotypeCaller to emit per-sample germline variant calls or gVCFs from analysis-ready BAM/CRAM inputs.

Use when splitting mixed accession-like text into one lowercase token per line in EDirect-style text pipelines.

Use when converting ACE assembly files into SAM while preserving legacy ACE-specific padded or contig-sequence behavior.

Use when pretty-printing tab-delimited output with left, center, right, or decimal-aware numeric alignment in EDirect text workflows.

Use when masking columns or coordinate ranges in multiple-sequence alignments before downstream HMMER or alignment-processing steps.

Use when counting amino-acid letters in raw protein sequence lines inside simple EDirect text pipelines.

Use when passing legacy ViennaRNA distance-matrix data through the AnalyseDists helper for simple formatting or downstream plotting workflows.

Chinese-context AI news voice. Use this agent when a global AI event should be translated into practical consequences for Chinese-speaking creators, teams, and product builders.

Research-first AI news voice. Use this agent when an event should be interpreted from the perspective of lab progress, capability changes, benchmarks, and what is actually new.

Builder-first AI news voice. Use this agent when an event should be interpreted for developers, operators, or tool builders who care about integration, usability, and ecosystem impact.

Product-and-market AI news voice. Use this agent when an event should be interpreted for product teams, founders, operators, and people tracking commercial direction.

Editorial-routing AI news voice. Use this agent when an event should be interpreted as part of a broader information map across feeds, sources, and repeating signals.

Safety-and-governance AI news voice. Use this agent when an event should be interpreted through risk, alignment, rollout safety, and broader externalities.

Critical AI news voice. Use this agent when an event needs a careful read of incentives, missing evidence, rollout risk, and narrative inflation.

把 dsh 插件做成可安装/可发布的包时使用——package.json 的 dsh.bundle / dsh.profile / dsh.client 字段、cordis.patch.yml 写法、patch 层的应用顺序与按 id 覆盖、dsh plugin add（tgz / git / npm）、profile 目录结构、从 GitHub 安装的 prepare + allowBuilds 坑、files 白名单与 publint、发布与版本同步。也用于排查"插件装了但没生效"。

在回答或实现任何 DeepSeek Harness / dsh / Cordis 上游行为前使用——选择扩展点、查 ctx 服务或事件、确认配置项、工具 schema、生命周期、profile/bundle、Client API，或判断文档是否适用于插件当前版本。先核对 upstream lock 与插件依赖，再读 vendor/deepseek-harness 中同一版本的源码文档和实际 .d.ts；也用于克隆、刷新和审阅上游版本。

要在真实 DeepSeek Harness 安装上读取、安装、启动或验证插件时使用——任何会碰 $DSH_HOME / ~/.dsh、profile、dump-config、tarball 安装、Web UI、Session 日志或回滚的操作都先用它。保护凭据、settings 写者锁、用户 patch、真实会话与持久数据；按只读配置检查→隔离 overlay→tarball→启动的代价顺序验证，并在结束时报告所有环境改动。

写或改 DeepSeek Harness（dsh）插件本体时使用——新建插件、加 Config、提供或消费 Cordis 服务、挂事件监听、处理生命周期与资源清理、给 ctx 加类型、决定拆几个包、给服务/类取名。涉及 apply/inject/ctx.effect/ctx.plugin/Service 基类/waterfall 等概念时都适用。不覆盖工具定义（见 dsh-tool-authoring）和打包安装（见 dsh-bundle-and-profile）。

想知道某个 DeepSeek Harness 插件是否已经存在、寻找成熟实现、盘点当前工作区插件、或评估第三方 dsh 插件能否安装时使用。写新插件前必须先查；用 tools/find-dsh-plugins.py 核实 npm/GitHub 候选的 package.json dsh 字段、版本与源码可信度。涉及本机已装插件时只读发现并把清单留在被忽略的 my-plugins.md，不公开真实 profile 或用户数据。

改完 dsh 插件代码、准备提交或发布之前使用——上游到底要求什么、该跑哪些门禁、门禁缺失怎么识别、生成物哪些该入库、客户端半该用插槽还是 DOM 注入、提交前清单。流程属于这个工作区，不往插件仓库里塞 CI；跑门禁用 tools/check-plugin.py。

写或改 dsh 里模型可调用的工具时使用——defineTool、parameters/output schema、execute 契约、exec.signal/exec.agent、后台任务（run_in_background + ctx.jobs）、UI 卡片（presentCall/presentResult/presentationMeta）、以及该用哪个执行策略钩子（tools/pre-execute、ctx.tools.guard、tools/execute、tools/post-execute、too…

Edit an existing blog post. Shows current content, applies user-requested changes, and saves.

List all blog posts with metadata. Shows titles, dates, tags, and available languages.

Create a new blog post. User provides a topic or title, Claude generates content and creates the post.

Build, commit, and deploy the blog to GitHub Pages.

Understand and customize the local Trellis architecture inside a user project. Use when modifying .trellis plus platform hooks, settings, agents, skills, commands, prompts, or workflows generated by trellis init.

Write long-form answers as high-density newspaper-style pages when the user wants a report, brief, feature page, morning brief, project update, or highly scannable HTML/editorial output. Use for long replies that should be easy to scan first and deep-read…

Port, reconstruct, modernize, or behaviorally reimplement Windows, macOS, and legacy applications as high-quality Linux applications. Use for executable or installed-application analysis, source recovery, platform adaptation, Linux rewrites, Electron or .NET…