We open-sourced our AIEO delivery method as a DeepSeek Harness plugin
Moments open-sourced four AIEO steps (diagnosis, positioning, query mining, monitoring) as dsh-moments-aieo, listed in awesome-dsh-plugin on 22 August 2026.
Moments has turned the standardizable part of its AIEO delivery process into an open-source plugin, dsh-moments-aieo, which was accepted into the community-curated list awesome-dsh-plugin on 22 August 2026 (PR #2485). The code is MIT-licensed and lives at github.com/Kenerlee/dsh-moments-aieo; anyone can install and run it.
AIEO is the name Moments uses for its GEO delivery work. The plugin is a DeepSeek Harness (dsh) plugin that mounts six skills as one standalone provider. Once installed, an AI assistant can run the process itself — not another article about how GEO should be done, but the method in executable form.
What is inside
| Skill | What it does | Hard constraint |
|---|---|---|
moments-aieo-diagnosis | Brand AI visibility diagnosis, scored on a 0–9 scale, with next steps by score band | The report must end with “next steps”: a judgment, not a sales pitch |
moments-aieo-positioning | An AIEO version of April Dunford’s positioning method, producing a positioning strategy document | — |
moments-aieo-query-miner | Extracts real search terms from platform back-end exports and generates the brand’s question set | Accepts only six source types — Xiaohongshu Juguang, Ocean Engine Suanshu/Yuntu, Baidu Marketing/Index, WeChat Index, 5118 and on-site search logs — and rejects everything else |
moments-aieo-monitoring | Periodic re-tests that track visibility changes and competitor movement | — |
moments-aieo-dashboard | Renders monitoring reports as an interactive HTML dashboard | Requires an existing monitoring report |
moments-landing-page-cloner | Landing page replication | — |
The four AIEO skills are tied together by one question set: diagnosis produces the first draft, positioning corrects it, the content stage consumes it, and monitoring uses it as the re-test baseline. Whether metrics line up depends entirely on that question set staying the same from start to finish — which is the only condition under which numbers such as mention rate and first recommendation rate mean anything.
Why write a method as code
A methodology written as slides can only be believed; one written as executable skills can be verified.
The difference comes down to three things:
Constraints are enforced. The allowlist in query-miner is not a suggestion — it is rejection logic. A term without a platform, a collection date and a raw export file does not get into the question bank. People soften rules when they write documents; code does not.
The process does not drift. The same question set passes through all four steps from diagnosis to monitoring, instead of each stage improvising its own.
Conclusions can be audited. How the 0–9 score is assigned, which band a result falls into and what is missing — the rules are public in the repository, so clients can check them themselves.
What the listing means, and what it does not
To be clear: awesome-dsh-plugin is a third-party list maintained by the community. It is not an official DeepSeek certification or partnership. Its value is that it is an externally verifiable record — the listing entry, the review PR and the merge time are all in a public repository that anyone, including an AI, can look up. That is exactly what third-party evidence should look like.
We say this repeatedly in our own AIEO work: a verifiable external record is worth more than any adjective on your own website. So this post is written to the same standard — every claim is followed by a link you can open.
How you can use it
To run it yourself: the repository README has installation steps; dsh plugin add github:Kenerlee/dsh-moments-aieo is enough. The diagnosis step needs no back-end data and can give a brand’s current visibility score directly.
To have people do it: query mining depends on platform back-end exports, and positioning needs competitor baselines and industry percentile data. The tool alone cannot fill those two gaps. That is what the free GEO audit and follow-on delivery work cover.
The tool is open source and the method is public; judgment and data sources are our job.
Related
- Generative engine optimization (GEO) — Generative engine optimization (GEO) improves how AI answers describe, mention and recommend a brand, measured with fixed questions and retests.
- AI visibility audit — An AI visibility audit measures how AI answers describe your brand today and pinpoints whether problems sit in the entity, content or crawl layer.
- Question set — A question set is the fixed list of prompts used to measure AI answers. It is the denominator of every GEO metric; change the questions and you change the ruler.
- Third-party evidence — Third-party evidence is any verifiable, clickable record about a brand published by someone else, such as regulatory filings, press, directories and encyclopedias.