dsh-moments-aieo
Manifest validAIEO (GEO/AEO) delivery method as a DeepSeek Harness bundle: five moments-aieo-* skills over a shared question library

dsh-moments-aieo
English | 中文
What this method finds, on real data. In November 2026 we asked ChatGPT (with search) 40 English buyer questions about Chinese mining machinery suppliers, twice. SBM was named in 50% of the 80 answers, and 77% of the links ChatGPT cited pointed to the manufacturers' own websites. Read the study · questions, scores and script
An AIEO (AI Engine Optimization — the GEO/AEO practice of getting a brand cited by ChatGPT, DeepSeek, Doubao, Kimi, Perplexity and friends) delivery method, packaged as one DeepSeek Harness bundle. The method runs in four stages — diagnosis → positioning → content → monitoring — chained by one question bank: diagnosis drafts it, positioning corrects it, content consumes it, monitoring measures against it. This bundle ships the three stages that are method rather than writing — diagnosis, positioning, monitoring — plus the question bank itself, as one named skill provider. The content stage consumes the bank through whatever writing skill you already use.

Quick start: Claude Code, Codex and other Agent Skills tools
Every skill is a standard SKILL.md folder, so you don't need DeepSeek Harness to use them. Copy the whole skills/ directory, including shared/, which holds the question bank the skills read through ../shared/:
git clone https://github.com/Moments-top/dsh-moments-aieo.git
cp -R dsh-moments-aieo/skills/. ~/.claude/skills/ # Claude Code
# other tools: copy into their skills directory, e.g. ~/.agents/skills/
Then ask in plain words, for example "run an AI visibility diagnosis for read, glob, web_fetch); agents map them to their own tools.
Plugin
Requires ctx.skills (inject: ['skills']).
The plugin body is deliberately thin: it mounts @deepseek-ai/dsh-skill-filesystem with includeDefaultRoots: false over its own skills/ directory, so this set registers under one provider name and never collides with same-named skills in ~/.dsh/skills or ~/.agents/skills. No scanner, watcher, or frontmatter parser is reimplemented here.
Config
| Field | Default | Meaning |
|---|---|---|
skillsDir | the package's own skills/ | Directory holding the <name>/SKILL.md bundles. Point it at a working tree during development. |
providerName | moments-aieo | Provider name registered on ctx.skills, keeping this set separable from the user's own roots. |
Install
dsh plugin --profile web add github:Moments-top/dsh-moments-aieo # straight from GitHub
dsh plugin --profile web add file:/path/to/clone # from a local clone
Then add the package to the profile's bundle list in ~/.dsh/profiles/web/package.json:
{ "dsh": { "profile": { "bundles": [
"@deepseek-ai/dsh-base",
"@deepseek-ai/dsh-web-app",
"dsh-moments-aieo"
] } } }
The bundle's own cordis.patch.yml inserts the row, so no profile patch is required. Override it by id in ~/.dsh/profiles/web/cordis.patch.yml when you want your own skill directory:
- id: moments-aieo
config:
skillsDir: /absolute/path/to/your/skills
Verify without booting:
dsh --profile web --dump-config | grep -A 4 'id: moments-aieo'
Browser automation
moments-aieo-diagnosis collects real answers through bb-browser (install it and log in to each AI platform in its managed browser); it renders a visual client report plus an evidence zip with stdlib-only Python scripts. moments-aieo-monitoring drives the platforms through Playwright. dsh reaches MCP servers through dsh-mcp-client, which registers their tools under mcp__<serverName>__<rawName> — the same server-qualified shape Claude Code uses, so the mcp__playwright__browser_* names in the monitoring skill resolve as long as the server is named playwright:
- insert:
- id: mcp-playwright
name: '@deepseek-ai/dsh-mcp-client'
config:
serverName: playwright
command: npx
args: ['@playwright/mcp@latest']
Without the browser tooling both skills still produce a website audit and the question bank; the platform-visibility measurements are what go missing.
Screenshots
A diagnosis report and the monitoring dashboard, both from real client runs with the brand redacted.

Skills
| Skill | Purpose |
|---|---|
moments-aieo-diagnosis | Brand AI-visibility diagnosis; emits a visual client report (snapshot, competitor visibility, question × brand matrix, fact accuracy, sources, 90-day roadmap), an evidence zip and the first draft of the question bank |
moments-aieo-positioning | Positioning analysis on an AIEO-adapted April Dunford method; iterates the question bank |
moments-aieo-query-miner | Real search-term mining from whitelisted platform exports only; refuses to invent terms |
moments-aieo-monitoring | Periodic visibility, share-of-voice, content-quality and conversion tracking |
moments-aieo-dashboard | Renders monitoring reports into an interactive HTML dashboard |
moments-landing-page-cloner | High-fidelity landing-page replication |
Diagnosis, positioning, query mining and monitoring share one artifact chain: the question bank the diagnosis drafts is what positioning corrects, content consumes, and monitoring measures against. Running them out of order is allowed and produces a weaker bank.
Model Experience
Indirectly, through @deepseek-ai/dsh-tool-skill: this provider's names and capped descriptions appear in the model's skill catalog, and skill(name) loads the selected SKILL.md body plus its resource base. Paths, provider ranks, and the mount configuration stay hidden from the model.
KV Cache effect
Catalog only. Registration adds eight rows to the catalog digest once; skill bodies enter history only when the model loads one.
Known Limitations and Deferred Work
- The frontmatter
allowed-toolskey does nothing under dsh — the parser readsname,description,whenToUse,metadataand the two invocation flags, and ignores the rest. It neither errors nor restricts anything; the key is kept for Claude Code compatibility. Harness tool names in the bodies were corrected to dsh spellings (read,glob,web_fetch); MCP names need the server configured above. - Web mode disables the host-level provider —
dsh-web-appsetsskill-filesystem: disabledbecause agent presets own local discovery. This bundle registers globally and preset agents read the merged catalog, so the set stays visible; a deployment that isolates its presets from global registrations would not see it. - No build step — the plugin ships as plain
.mjswith no TypeScript source, nolib/, and no type declarations. It is twenty lines; a consumer wanting types writes them. - Reference cases are not distributed — the diagnosis skill's worked client examples live outside this repository.
- Chinese-first content — every AIEO skill body is written in Chinese, and the scoring rubrics assume Chinese-language AI search platforms.
Who built this
The method comes from real AIEO delivery work — brand diagnosis, positioning, question-bank construction and monitoring for consumer, healthcare, SaaS and franchise clients. It measures mention rate and first-recommendation rate over a frozen question set; dated example data is in Brand visibility in Chinese AI assistants. The tooling is open source; the industry baselines and the judgement of what to do with a low score are not things a Markdown file can carry. Want a human-run baseline instead? Free AI visibility audit.
Ran a diagnosis? Open a Discussion with your score and industry (no brand name needed). Real numbers across industries are what turn a scoring rubric into a benchmark, and the aggregate goes back into this repo.
License
MIT
Comments
Loading…
Similar plugins
by Azzygoatcoder
模块化 AI 科研/工程技能 monorepo(DeepSeek Harness / Claude Code 通用)— plugins/ + skills/ + bin 脚本 + LaTeX 模板,验证环驱动
★ 9
MIT
Python
Sep 24, 2026
dsh plugin --profile web add agent-useful-skillsby dongsheng123132
Content-addressed receipts for skills actually loaded by DeepSeek Harness
★ 3
MIT
JavaScript
Nov 6, 2026
dsh plugin --profile web add dsh-capability-receiptInstall SandBase Agent Skills into DeepSeek Harness projects.
★ 54
Apache-2.0
Python
Aug 23, 2026
dsh plugin --profile web add @sandbaseai/dsh-skillsby liangdabiao
dsh-geo-optimizer · GEO 工具箱 DeepSeek Harness(dsh)插件:生成式引擎优化(GEO)三个技能——让 ChatGPT、豆包、Perplexity、文心一言这些 AI 在回答问题时主动提到你。GEO toolkit dsh plugin: 3 skills for generative engine optimization
★ 0
MIT
Python
Nov 9, 2026
dsh plugin --profile web add dsh-geo-optimizerby xuboboo
DSH (DeepSeek Harness) plugin bundle: grok-geo GEO brand audit agent skill - AI-search visibility, recommendations, citations, competitor presence and content-gap diagnosis across 17+ AI engines, ship
★ 3
MIT
Python
Aug 14, 2026
dsh plugin --profile web add dsh-grok-geoby PerryLink
Double-check before you ship: grill the requirements, test the implementation, prove the delivery. An engineering-discipline bundle for DeepSeek Harness.
★ 58
↓ 1.2k/wk
Apache-2.0
TypeScript
Nov 13, 2026
dsh plugin --profile agent add dsh-doublecheck