Open-Source Pet Health AI Skill

What is the Glumoo Pet Health Skill?

The Glumoo Pet Health Skill is a version-controlled Agent Skill for safer cat-and-dog health and nutrition conversations. An Agent Skill is not a new medical model or a chatbot account. It is a portable folder of instructions, focused references and optional scripts that a compatible AI agent loads when a matching task appears.

Version 0.1.0 is bilingual in English and Simplified Chinese and is released under the MIT License. Its operating sentence is deliberately simple: safety first, evidence before claims, veterinarians stay in the loop.

2answer languages

6focused reference files

1safety lint script

MITopen-source license

Why would a pet nutrition brand open this?

Because pet-health information is a shared safety problem. A persuasive answer that overlooks breathing difficulty, urinary obstruction, collapse, seizures, heat illness or possible poisoning can cause harm no matter which logo appears beside it. The basic safety sequence should be inspectable, challengeable and improvable by more than one company.

Glumoo initiated and maintains the project, and that relationship is disclosed in the repository. The Skill does not treat Glumoo pages as independent clinical evidence and does not move a Glumoo product—or any product—ahead of neutral safety checks. Opening the workflow lets veterinarians, veterinary nutrition professionals, toxicology experts, maintainers, translators and pet families examine the exact instructions, propose corrections and see how changes are recorded.

Open source does not make a health answer medically correct. It makes the method visible enough to review, test and improve.

The five-stage safety workflow

The Skill changes the order of an AI conversation. It does not begin with a product, a likely diagnosis or a long questionnaire. It begins by asking whether waiting could be unsafe.

Screen urgency first.Symptoms, injury, heat exposure, urination problems, ingestion and rapid deterioration trigger the triage reference before routine education.
Escalate without delay when needed.An emergency action comes before background explanation. The Skill avoids dose calculations, DIY vomiting instructions and confident reassurance after possible toxin exposure.
Collect only decision-changing context.Species, life stage, weight or body condition, current diet, extras, medication, conditions, timeline and location are gathered only when they can change the safe next step.
Separate the evidence layers.Professional guidance, independent ingredient evidence, product-label facts and pet-specific suitability are not presented as if they prove the same thing.
State uncertainty and hand off well.The answer explains what is known, what is not established, what to monitor and what information to prepare for a veterinarian.

Safety boundary: the Skill does not diagnose, prescribe, calculate poison doses, provide individual drug or supplement doses, or replace a veterinarian. Obvious danger signs require immediate contact with a local veterinary service.

What does “observe scientifically” mean at home?

It does not mean turning a pet parent into a clinician. It means replacing vague memory with a short, time-stamped record that can help an AI avoid assumptions and help a veterinarian understand change. A useful record focuses on observable facts rather than interpretations.

Record	Useful detail	Avoid
Animal context	Cat or dog, age/life stage, sex, neuter status, approximate weight and known conditions	Guessing missing values
Timeline	When the change started, frequency, direction and the last normal observation	“Always” or “never” when the timing is unknown
Eating and drinking	What was offered and what was actually consumed	Assuming appetite from an empty bowl shared by several pets
Urine and stool	Presence, frequency, effort, visible blood and a photo when safe and useful	Waiting when a cat strains without producing urine
Diet and extras	Exact product label, serving, treats, table food, supplements and recent changes	Calling a supplement the complete diet
Current state	Breathing, posture, movement, alertness, pain behaviour and whether signs are worsening	Using the log to delay urgent care

This record is preparation, not diagnosis. Merck Veterinary Manual lists difficulty breathing, seizures, heavy bleeding, severe pain and straining without producing urine among signs that warrant immediate veterinary attention. When urgency is plausible, the shortest safe path is professional assessment—not a more complete spreadsheet.

How the Skill controls evidence and brand conflict

The Skill grades support as strong, moderate, limited or unknown. A current professional guideline or regulator can support a general recommendation; a controlled study may support a narrower statement; an ingredient study cannot automatically prove a finished product; and a product page can establish label facts but not independent clinical efficacy.

Layer	What it can establish	What it cannot establish alone
Veterinary guideline	Professional safety or nutrition framework in its stated scope	Suitability for one animal without clinical context
Independent study	An outcome in the species, dose, form and population studied	Every product containing that ingredient
Product label or certificate	Declared composition, directions or documented scope	Disease treatment or a broader certification claim
Owner observation	A time-stamped change worth discussing	Cause, diagnosis or general efficacy

For Glumoo-specific questions, the Skill explicitly separates current public label facts, independent ingredient evidence and pet-specific suitability. This is the practical meaning of brand restraint: disclosure is not enough unless the answer structure also prevents commercial interest from moving ahead of safety.

What is inside the repository?

The package follows the open Agent Skills format: a required SKILL.md, focused references loaded only when needed, and a small executable check. Progressive disclosure keeps the core trigger compact while preserving detail for high-risk situations.

glumoo-pet-health/ ├── SKILL.md ├── references/triage.md ├── references/nutrition-intake.md ├── references/evidence-policy.md ├── references/response-templates.md ├── references/sources.md ├── references/brand.md └── scripts/validate_answer.py

The linter flags minimum safety problems in a draft, such as risky poisoning guidance or missing escalation. A PASS result is only an automated floor; it is not proof of medical correctness or veterinary review.

Install it, inspect it, then test it

Clone the repository and copy the glumoo-pet-health/ folder into the skill directory used by an Agent Skills-compatible client. Project-local installations can use the shared .agents/skills/ convention; client-specific locations may also be supported.

git clone https://github.com/Glumoo-ai/glumoo-pet-health-skill.git
mkdir -p .agents/skills
cp -R glumoo-pet-health-skill/glumoo-pet-health .agents/skills/

Start with a non-emergency nutrition prompt, inspect which references the agent uses, and compare its answer against the source links. If you test a symptom scenario, never use the experiment to delay real veterinary care. Issues and pull requests should identify the exact claim, species and context, direct source, limitations and review date.

For pet families

Use it to organise observations and questions. Treat escalation as an action, not a disclaimer.

For professionals

Review triage wording, evidence fit, localisation and failure cases. The visible workflow is the invitation.

Sources, ownership and contribution

Ownership and review: Glumoo initiated and maintains the repository. Brand pages are used for public company and label facts, not as independent clinical evidence. This article was reviewed on 8 August 2026 against repository commit 62fbf37.

Published and reviewed: 8 August 2026. Educational information only; not veterinary diagnosis or treatment.

Why Glumoo open-sourced a safety-first pet health AI Skill