Dog Head Strategist · Goutoujunshi
First validates emotions, then analyzes relationships and gives actionable steps—an AI relationship advisor that's truly on your side.
Evidence shows: README and SECURITY.md state memory script requires user consent and provides revocation and hard delete; data flow transparent (local SQLite, no chat database access); external effects limited (no server, no network service). Deductions: dependency security only mentions ChatLab as optional, no dependency list or vulnerability scan; source attribution only copyright notice, no author identity or contributor info.
Evidence shows: README and CHANGELOG consistent, workflow clear; dependency availability: ChatLab optional but no degradation behavior documented; failure messages: no error handling or user prompt examples. Deductions: lack of failure scenario documentation.
Evidence shows: explicit support for diverse sexual orientations, gender identities, and relationship scenarios; capability boundaries clear (not substitute for therapy); trigger phrase explicit ($goutoujunshi); environment fit: installation to Codex Skills directory. Deductions: no other environment compatibility mentioned.
Evidence shows: clear project structure, specific installation instructions, stable naming, examples and FAQ (typical answer flow), known limitations (not substitute for professional services), license clear (PolyForm Noncommercial), CHANGELOG present. Deductions: maintenance responsibility not explicit, no contributor guide or maintainer info.
Evidence shows: high output usability (provides ready-to-send phrases and action plans), clear marginal value (distinct from simple advice), reasonable cost-benefit (on-demand knowledge loading). Deductions: no actual run verification, cost-benefit is estimated.
Evidence shows: distinguishes facts, inferences, and unknowns; knowledge base distinguishes research evidence and popular claims. Deductions: no specific citations, insufficient cross-source corroboration.
- Dependency security information insufficient, no dependency list or vulnerability scan.
- Maintenance responsibility unclear, no contributor guide or maintainer info.
- No actual run verification, cost-benefit is estimated.
What does this agent do, and when should you use it?
Goutoujunshi is a Codex skill that functions as a relationship advisor, integrating emotional support, relationship science, and actionable strategies. It defines core behavior in SKILL.md, uses a multidisciplinary knowledge base in references/knowledge and references/practical (covering psychology, law, sociology, philosophy, etc.), and supports user-consented local long-term memory via scripts/memory_store.py with undo and delete capabilities. It can analyze chat screenshots, exported texts, and ChatLab data, identifying speakers and separating facts from speculation. Its output includes ready-to-send messages, action plans, observation windows, and stop conditions. Design principles emphasize processing emotions first, relying on behavior over labels, reciprocity over conquest, explaining costs, non-negotiable consent and exit rights, and safety first in dangerous situations.
Goutoujunshi receives user narratives, chat screenshots, or exported files and executes the workflow in SKILL.md: first grounds emotions, then profiles the user and target (MBTI, ratings, etc.), separates facts from speculation, retrieves relevant knowledge documents from references based on context, evaluates reciprocity, risk, and opportunity cost, and outputs a primary recommendation, phrases, action steps, observation windows, and stop conditions. It can invoke scripts/memory_store.py for long-term memory storage (only after explicit user consent), supporting viewing, pausing, undoing, and clearing; scripts/validate_skill.py checks project integrity. If ChatLab is configured, it can query chat logs to analyze relationship trends.
- A user wants to know how to reply to a specific message and gets a ready-to-send phrase with follow-up branches.
- A user wants to pursue, reconnect, or advance a relationship and gets a concrete action plan that adjusts based on feedback.
- A user is torn between multiple people and receives a clear judgment based on behavior, personality, and practical conditions.
- A user provides chat screenshots or exported files and gets automatic speaker identification and fact/speculation separation.
- A user wants to avoid repeating background across tasks and enables local long-term memory after first consent, with automatic updates and on-demand recall.
- A user faces conflict, imbalanced investment, or breakup choices and gets emotional validation, pros/cons analysis, next steps, and stop conditions.
What are this agent's strengths and limitations?
- Comprehensive multidisciplinary knowledge base covering psychology, law, sociology, philosophy, and separates research evidence from popular claims.
- Validates emotions before analyzing, fitting real counseling needs.
- Supports diverse relationships (LGBTQ+, non-monogamy) without gender role assumptions.
- Provides immediately executable phrases and action plans with stop conditions.
- Long-term memory reuses profiles across tasks, but emphasizes user consent and revocability.
- Currently only supports Codex platform; other environments need adaptation.
- Requires detailed personal and relationship info to fully exploit capabilities.
- Knowledge base relies on community contributions; updates may lag.
- Complex scenarios like legal or safety issues may need professional intervention; tool is not a substitute.
- Long-term memory and ChatLab are optional but require extra setup.
How do you install or deploy this agent?
Clone the repository into Codex's Skills directory:
git clone https://github.com/powerycy/goutoujunshi.git ~/.codex/skills/goutoujunshiNo additional credentials required.
How do you use this agent?
In Codex, enter:
使用 $goutoujunshi 帮我梳理情绪、分析当前关系，并给出下一步建议。On first use, it will ask a series of questions (e.g., MBTI, ratings, relationship status) that you can leave blank or tell your story. If ChatLab is installed and export files are ready, you can say:
使用 $goutoujunshi 通过 ChatLab 分析我和对象A最近三个月的聊天。Supports Chinese interaction.