Info: Machine translation This post was machine-translated from the Chinese original. Wording may be rough in places — the Chinese version is authoritative.
When receiving a baffling message, the real difficulty is often not “how to respond,” but first figuring out what the other person has done, what you want, and whether continuing to invest is worthwhile. Open-source Codex Skill “Dog-Head Strategist · Goutoujunshi” targets exactly this process: first process emotions, then break down facts, speculations, and unknowns, and finally provide actionable next steps.
It delivers not a one-size-fits-all response
According to the project documentation, the dog-headed strategist covers relationship scenarios such as heartbeat, ambiguity, invitations, conflict, investment imbalance, breakups, reconciliation, and marriage and family. The output can be a direct message, an action plan, observation window, pause conditions, or dialogue drills; When faced with multiple choices, it first analyzes separately, then compares reciprocity, real-world conditions, risks, and opportunity costs.
This approach is more like asking a friend to help organize a long string of WeChat messages: first mark what the other person clearly said, and which are just your own guesses, then decide whether to reply, invite them, wait, or leave. The difference is that Skill writes this sequence as a fixed process, avoiding reading someone’s mind immediately after seeing an expression or a cold reply.
- I just want to reply with one sentence: first give a replicable product, then explain the timing of sending and the follow-up based on different feedback.
- Chat records are long: distinguish the speaker, the visible original text, chronological order, and unknown information, without imagining text records as offline behavior.
- Imbalance in relationship investment: Lumping reciprocity, reliability, feasibility, and exit costs into the same judgment.
- If there is a risk of explicit denial, violence, stalking, coercion, or self-harm, stop normal advancement and prioritize safety and assistance.
Why this project deserves to be on the shortlist
During data collection, warehouse records were 7,350 stars and 740 forks; Baseline snapshot was 7,238 stars, 112 stars added in 1.02 days, recording speed about 109.73 stars/day, and the most recent push was 18.0 days after collection. The warehouse is not fork, nor is it archived or disabled; the license is marked as MIT. The above figures are snapshots of two time points and cannot replace actual measurements of functionality, security, or long-term maintenance quality.
The project also compiles 135 references and maintains topics such as relationship science, communication strategies, legal safety, marriage, family, and sexology separately. The document claims it will load only necessary information for the current question, and clearly treats MBTI, attachment types, and internet scripts as auxiliary questions and generation strategies, rather than diagnosis, fate judgment, or winning formulas.
How to get started, and what does it require?
The project is designed for the Codex Skill scenario. The author suggests starting by having the AI assistant install the native Skill from the public repository, then raise specific relationship questions. The repository entry is https://github.com/shengjidaguai-china/goutoujunshi. This is the installation path declared by the author and does not mean all chat clients that support custom commands are compatible.
When used for the first time, Skill will ask about the basic information of the user and the target person, relationship history, goals, and current emotions; Items unknown can be left blank. When saving streamlined files across tasks, documents require clear consent first and the ability to view, pause, revoke, or delete. When involving real names, chat content, and intimate relationship details, anonymization should still be done first, providing only necessary information that may affect judgment.
ChatLab is an optional dependency used to analyze chat export files that users have obtained and prepared. Doghead Strategist itself does not directly read, decrypt, or export the database of chat software such as WeChat or QQ; Without ChatLab, it can still paste text or upload screenshots. In other words, it can organize the “packages” delivered to you, but it will not break into the warehouse to pick up the items for users.
The free code is not the entire process that is already packaged
The repository uses the MIT License, allowing use, modification, distribution, and commercial use, but must retain copyright and license statements at the time of distribution. The license also clearly states that the software is provided as-is, without any guarantees. The repository does not prove that Codex, model calls, ChatLab, or other runtime environments are all free, so “code open source” cannot be written as “zero-cost throughout use.” Actual fees and account conditions should be subject to the rules of the selected client and service at the time.
The documentation does not list the standalone cloud services that must be purchased, nor does it provide a unified tables for models, systems, or hardware compatibility. Whether installation is possible, local files can be called, image input is supported, and where long-term memory is stored all depend on the client, permissions, and runtime environment used, so you need to confirm these before installation.
The author’s explanation of what can be done does not mean it has already been independently executed
Current data verification covers the repository’s README, SKILL.md, licenses, and code trees, and runtime validation has not yet been performed. Therefore, it can be confirmed that the project has written the installation entry point, analysis process, optional dependencies, memory consent mechanism, and security boundaries; It cannot be claimed based on this that response quality, cross-client compatibility, privacy implementation, or long-term memory erasure have passed independent testing. The process and instructions shown by the author should not be passed as the user’s own running results.
Warning It provides relationship education and decision support, not psychological therapy, medical diagnosis, legal advice, police or emergency services. When facing domestic violence, stalking, coercion, financial control, self-harm, or personal danger, don’t treat ordinary love phrases as a solution.
Who is worth using?
- Suitable for: Those who have already used Codex, are willing to separate facts, guesses, and goals, and want to obtain specific actions and stop-and-stop conditions.
- Suitable for: Those who need to organize longer chat logs but are willing to first confirm the speaker, time frame, and privacy boundaries.
- Not suitable for now: expecting a single sentence to guarantee winning someone’s heart, or hoping to read someone’s mind with a label.
- Not suitable for now: people unwilling to hand sensitive data over to current clients, and who have not confirmed local storage, image uploads, and model service boundaries.
A reliable way to accept is to first test three things with anonymized, low-risk old dialogues: whether it can correctly distinguish facts from speculation, whether it respects explicit refusal and withdrawal rights, and whether the next steps are specific and allow for regression. After these checks, decide whether to provide more background or enable long-term memory. The most valuable part of the dog-headed strategist is not guessing the answer about love for others, but relaying out the evidence, choices, and costs that easily mix when emotions run high.
Project firsthand information
🧰 Tools I build
I maintain all of these tools myself. Preview builds are clearly labeled; the release pages are the source of truth for downloads, updates and known limits.
Info: HyphenBox Status: Official releases
A radar for free LLM APIs: availability is re-tested continuously, one local interface for all of them, and keys stay on your machine
Info: LocalBrain Status: Official releases
A multimodal MCP toolbox for local models: TTS, Whisper and video generation in one place
Info: ScreenLex Status: Official releases
Learn new words while you watch shows. Free, for Mac and Windows
Info: HyphenScreen Status: Official releases
Screen recording and smart editing in one: a DaVinci-style timeline, automatic redaction and a check of the finished video before export. Free
Quote: HyphenTech Make AI your superpower Local deployment · Free resources · Self-made software https://hyphentech.top
Late nights and burned API credits went in,a cup of tea comes back out — only if you feel like it.
Scan with WeChatPress and hold to save the image, then open it from your album in WeChat Scan
Comments
Loading comments…