倪海厦skill · 经方中医AI

• Outcome: Review
• Latest audit: May 24, 2026, 2:48 PM
• Version:
  1.1.2

Overview

This is a markdown-only traditional Chinese medicine skill, but it needs Review because it can steer users toward individualized diagnosis, formulas, dosing, toxic substances, invasive procedures, and emergency-care alternatives without strong enough safety gates.

Install only if you intend to use it as historical or educational TCM reference and your host agent can enforce medical-safety limits. Do not rely on it for diagnosis, prescriptions, dosing, toxic substances, emergency symptoms, cancer, stroke, heart disease, kidney failure, pregnancy, or medication decisions; those require licensed medical care.

SkillSpector (32)

By NVIDIA

Context-Inappropriate Capability

High

• Confidence: 96% confidence
• Finding: The skill contains extensive disease diagnosis, triage, and treatment instructions, including condition-specific remedies and dosing-style guidance, while presenting itself as an authoritative medical persona. In an agent setting, this creates a high risk of unsafe medical advice, delayed real-world care, and harmful self-treatment, especially for serious conditions like stroke, cancer, and infection.

Intent-Code Divergence

Medium

• Confidence: 82% confidence
• Finding: The skill gives inconsistent instructions: it says answers must be grounded in retrieved internal source text, but elsewhere permits direct responses from an internal 'mind model.' That inconsistency can cause the agent to skip retrieval and present unsupported medical claims as authoritative, increasing hallucination risk in a high-stakes domain.

Context-Inappropriate Capability

High

• Confidence: 99% confidence
• Finding: The file is framed as a reference module, but it contains extensive diagnostic and treatment instructions for numerous conditions, including dosing, procedure selection, and claims of efficacy. In an unknown-purpose skill context, this can cause an agent to dispense actionable medical advice as if it were safe reference material, creating substantial risk of harmful self-treatment or delayed professional care.

Intent-Code Divergence

High

• Confidence: 98% confidence
• Finding: The module metadata says it is for syndrome differentiation and formula lookup, but the content is not passive lookup material; it actively instructs treatment choices and substitutions. That mismatch is dangerous because downstream systems may trust the file as low-risk reference content and expose unsafe medical guidance without stronger safeguards.

Intent-Code Divergence

High

• Confidence: 97% confidence
• Finding: The text claims the section can treat essentially all heart disease and presents highly confident efficacy statements inconsistent with a reference-style module. Such overclaiming can cause overreliance on unsafe, unvalidated advice and discourage urgent medical evaluation for potentially life-threatening cardiac conditions.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The README includes symptom analysis, formula-selection, and herb-use prompts that encourage users to seek individualized health guidance from the skill, but the only disclaimer appears later and is not placed near the examples. In a medical context, this can lead users to rely on the agent for diagnosis or treatment decisions, especially where the examples mention potent substances like aconite-derived preparations and sulfur, increasing the risk of harmful self-treatment or delayed professional care.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The skill exposes high-risk medical content without clear user-facing warnings that it is not a substitute for diagnosis or treatment by a licensed professional. Because it covers potentially life-threatening conditions and interventions, omission of safety warnings materially increases the chance that users rely on it for dangerous self-management.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The workflow explicitly routes diagnosis and treatment questions into direct answer generation, including condition matching and formula selection, without mandatory safety gates or referral prompts. In a medical context this is dangerous because it operationalizes unsafe advice delivery rather than merely describing historical or educational material.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: This file contains extensive, highly actionable medical diagnosis and treatment instructions, including symptom-based self-differentiation, named prescriptions, dosing concepts, preparation methods, contraindications, and emergency-style recommendations, but does not provide clear user-facing warnings to seek licensed medical care or avoid self-medication. In context, this is especially dangerous because the content is framed for real-world use in '感冒发烧初诊、方剂选择' and includes potentially hazardous substances and techniques (for example 麻黄、附子、半夏、攻下、发汗、火疗、针刺/放血), which could lead users to delay proper treatment, misuse potent substances, or harm vulnerable patients.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: This file contains extensive diagnostic and treatment guidance, including syndrome differentiation, contraindications, and named formulas, presented in an instructional style that could reasonably be acted on by users. Because it lacks a prominent warning that the content is educational only and not a substitute for licensed medical care, it may encourage self-diagnosis, self-treatment, delayed care, or unsafe use of potent substances and regimens.

Missing User Warnings

High

• Confidence: 93% confidence
• Finding: The text includes self-experimentation with prolonged forced breathing practice and physiological effects without any safety framing or contraindications. Even non-drug self-experiments can be risky for people with cardiopulmonary conditions, dizziness, or other vulnerabilities when presented as validation of medical theory.

Missing User Warnings

High

• Confidence: 96% confidence
• Finding: The passage gives specific prescription/substitution guidance for a syndrome and recommends an herbal formula combination without risk screening, dosage safety controls, or clinician oversight. In an agent skill, that is actionable self-medication guidance and may lead users to misuse herbs or delay proper care.

Missing User Warnings

Critical

• Confidence: 99% confidence
• Finding: The text explicitly suggests experimenting on a timid person by inducing sweating and then intentionally startling them to trigger illness. This normalizes non-consensual human experimentation and encourages deliberate induction of potentially dangerous symptoms, including cardiopulmonary distress.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The text recommends cold-water exposure for heart conditions as therapeutic advice without screening, contraindications, or safety warnings. Cold exposure can acutely worsen cardiovascular instability in vulnerable people and is especially dangerous when presented as a direct home treatment.

Missing User Warnings

Critical

• Confidence: 99% confidence
• Finding: This section provides handling and ingestion guidance for highly toxic substances such as乌头 in home-prepared remedies, including preparation details and dose-escalation style advice. In an agent context, this is dangerous because it facilitates poisoning, dosing errors, and serious cardiac/neurotoxic harm.

Missing User Warnings

Critical

• Confidence: 99% confidence
• Finding: The document explicitly recommends escalating to potent toxic substances when standard warming herbs seem ineffective, without any safety framework. This materially increases risk of severe poisoning and converts historical materia medica into actionable hazardous instruction.

Missing User Warnings

Critical

• Confidence: 99% confidence
• Finding: The passage discusses preparation and use of mercury-related materials like cinnabar/vermillion as treatment, including operational detail and normalization of ingestion/application. Mercury-containing substances present clear toxic risk, and such instructions are unsafe in any general-purpose agent skill.

Missing User Warnings

Critical

• Confidence: 99% confidence
• Finding: This section gives practical instructions for using poison-based preparations to kill parasites, including rationale for luring worms to ingest the toxin. It is directly actionable hazardous medical guidance with substantial risk of poisoning and misuse.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: This section gives explicit emergency stroke/coma interventions such as bloodletting, burning hair into powder and blowing it into the ear, and framing them as effective revival methods, without any warning that they are dangerous, unverified, and inappropriate outside licensed emergency care. A reader could delay real stroke treatment or directly injure a patient, causing death, brain damage, infection, aspiration, bleeding, or other severe harm.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: This section provides preparation and dosing guidance for highly toxic substances such as 乌头/附子, including escalation advice and claims that serious deforming disease can be reversed, while only lightly noting toxicity. That materially lowers the barrier to self-administration of a narrow-therapeutic-index poison and could lead to fatal poisoning, arrhythmia, neurologic toxicity, or delayed hospital care.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: This long section repeatedly recommends broad self-treatment for serious conditions such as lung abscess, tuberculosis-like illness, pulmonary edema/effusion, severe respiratory distress, and possible cancer, often with concrete formulas and comparisons implying that users can manage them directly. Without warnings about diagnostic uncertainty and urgent need for professional evaluation, readers may misdiagnose dangerous disease and postpone emergency or specialist care.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: This file contains extensive disease-treatment guidance, diagnostic claims, emergency care instructions, and specific herbal/drug dosing advice, while lacking any medical safety warning, contraindication guidance, or direction to seek licensed professional care. The content also includes dangerous anti-medical claims that discourage standard care (for example around vaccines, antihypertensives, cancer treatment, dialysis, biopsy, and diabetes management), which materially increases the risk of users delaying or rejecting evidence-based treatment and suffering serious harm.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: This file contains extensive, concrete medical treatment instructions for severe diseases including cancer, kidney failure, brain tumors, stroke, and lupus, with dosage-like details, treatment rules, and advice to avoid or replace standard medical care, yet it provides no safety warning, contraindications, or instruction to seek licensed medical supervision. In a skill context, users may treat this as actionable guidance, creating a serious risk of delayed evidence-based care, poisoning from dangerous substances, or other life-threatening harm.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: This section presents concrete treatment instructions, dosage-like guidance, and disease management advice as actionable medical content without an explicit warning that it is historical/educational material and not safe for self-treatment. Because the file repeatedly frames the advice as clinically effective and urgent, a user could rely on it instead of seeking qualified care, causing delayed treatment or direct harm.

Missing User Warnings

High

• Confidence: 99% confidence
• Finding: This portion includes invasive acupuncture and bloodletting guidance, including specific anatomical targets, depth/technique cues, and instructions around when not to bleed or where to puncture. Without strong safety warnings and professional-use restrictions, users may attempt dangerous self-procedures that can cause hemorrhage, infection, nerve injury, organ puncture, or death.

Static analysis

No static analysis findings were reported for this release.

VirusTotal

VirusTotal findings are pending for this skill version.