Back to skill

Security audit

Epidemiology of Infectious Diseases-PMPH-10edition

Security checks for vulnerabilities and agentic risk

Overview

The package is a Markdown-only infectious-disease skill set, but it contains detailed high-risk medical treatment guidance and bioweapon threat-classification content that should be reviewed before installation.

Install only in a controlled clinical or educational setting with system-level medical safety policy. Do not use it for self-diagnosis, self-medication, automatic prescribing, or emergency decisions without qualified clinician review, and restrict the bioweapon module to lawful public-health or biodefense use.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (251)

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · README_EN.md (reported line 62)May include surrounding context.

md
## Catégories de compétences

| Catégorie | Nb | Description |
|-----------|----|-------------|
| 🦠 Infections virales | 27 | COVID-19, Grippe, VIH/SIDA, Hépatite, Fièvre hémorragique, Encéphalite japonaise |
| 🧫 Infections bactériennes | 19 | Méningite, Typhoïde, Dysenterie, Choléra, Diphtérie, Anthrax, Peste, TB |
| 🌀 Spirochètes, Rickettsies & Champignons | 5 | Leptospirose, Syphilis, Lyme, Aspergillose, Cryptococcose |
| 🪱 Infections parasitaires | 15 | Paludisme, Ankylostome, Ascaridiase, Téniasis, Clonorchiase, Toxoplasmose, Kala-azar |
| 🐾 Zoonoses & infections spéciales | 2 | Brucellose, Botulisme, Classification des armes biologiques |
| 🏥 Contrôle des infections | 6 | Désinfection, Diagnostic IAS, Classification clinique |
| 🚑 Soins intensifs & gestion globale | 6 | Sepsis, Support d'organe, Gestion antimicrobienne, Sérologie |

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · README_FR.md (reported line 42)May include surrounding context.

md
## Catégories de compétences

| Catégorie | Nb | Description |
|-----------|----|-------------|
| 🦠 Infections virales | 27 | COVID-19, Grippe, VIH/SIDA, Hépatite, Fièvre hémorragique, Encéphalite japonaise |
| 🧫 Infections bactériennes | 19 | Méningite, Typhoïde, Dysenterie, Choléra, Diphtérie, Anthrax, Peste, TB |
| 🌀 Spirochètes, Rickettsies & Champignons | 5 | Leptospirose, Syphilis, Lyme, Aspergillose, Cryptococcose |
| 🪱 Infections parasitaires | 15 | Paludisme, Ankylostome, Ascaridiase, Téniasis, Clonorchiase, Toxoplasmose, Kala-azar |
| 🐾 Zoonoses & infections spéciales | 2 | Brucellose, Botulisme, Classification des armes biologiques |
| 🏥 Contrôle des infections | 6 | Désinfection, Diagnostic IAS, Classification clinique |
| 🚑 Soins intensifs & gestion globale | 6 | Sepsis, Support d'organe, Gestion antimicrobienne, Sérologie |

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · anthrax-clinical-treatment/SKILL.md (reported line 2)May include surrounding context.

md
---
name: anthrax-infection-control
description: 提供炭疽传染病的法定分类管理、隔离控制、接触者医学观察及疫区预防接种的标准操作流程。当用户询问炭疽病例的隔离标准、肺炭疽与皮肤炭疽的管理级别区别、密切接触者观察期,或疫区高危人群的炭疽疫苗接种方案时使用此技能。
---

# 炭疽传染病管理与预防控制

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · anthrax-clinical-treatment/SKILL.md (reported line 46)May include surrounding context.

md
---
name: anthrax-infection-control
description: 提供炭疽传染病的法定分类管理、隔离控制、接触者医学观察及疫区预防接种的标准操作流程。当用户询问炭疽病例的隔离标准、肺炭疽与皮肤炭疽的管理级别区别、密切接触者观察期,或疫区高危人群的炭疽疫苗接种方案时使用此技能。
---

# 炭疽传染病管理与预防控制

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · anthrax-infection-control/SKILL.md (reported line 2)May include surrounding context.

md
---
name: anthrax-infection-control
description: 提供炭疽传染病的法定分类管理、隔离控制、接触者医学观察及疫区预防接种的标准操作流程。当用户询问炭疽病例的隔离标准、肺炭疽与皮肤炭疽的管理级别区别、密切接触者观察期,或疫区高危人群的炭疽疫苗接种方案时使用此技能。
---

# 炭疽传染病管理与预防控制

Context-Inappropriate Capability

Critical
Category
Not specified by scanner
Confidence
99% confidence
Finding

This section provides explicit bioweapon-oriented threat assessment, including aerosolization suitability, lethality, environmental persistence, infectious dose, and mass-exposure handling. In a skill presented as a clinical infectious-disease resource, that capability is unjustified and directly increases misuse potential by helping users compare agents for offensive or harmful selection.

Content

No source excerpt is available for this finding.

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
97% confidence
Finding

This is a true harmful-content issue not merely because it names a dangerous pathogen, but because the surrounding section frames the agent in terms of weapon utility, lethality, and exposure scenarios. In the context of a clinical skill, such content is more dangerous because it embeds high-risk offensive knowledge in a tool that would otherwise be trusted for benign medical use.

Content

Scanner excerpt · bioweapon-pathogen-classification/references/pathogen-details.md (reported line 29)May include surrounding context.

md
- **人传人能力**:肺鼠疫型具有强人传人能力
- **环境稳定性**:较弱,对热和消毒剂敏感
- **作为生物武器特点**:可气溶胶释放引发肺鼠疫,传播迅速
- **处置要点**:链霉素/庆大霉素/多西环素有效,接触者需隔离观察

### 肉毒梭状芽孢杆菌 (Clostridium botulinum)
- **传播途径**:毒素摄入、伤口感染、吸入
- **致死率**:约5-10%(及时治疗),>50%(未治疗)
- **人传人能力**:无
- **环境稳定性**:芽孢极强,毒素在食物中稳定
- **作为生物武器特点**:毒素毒性极强,微量即可致死

Context-Inappropriate Capability

Critical
Category
Not specified by scanner
Confidence
99% confidence
Finding

The decision and response tables rank agents using criteria such as public panic, environmental stability, and especially large-scale production difficulty, which are not part of routine clinical care. These structured comparison tables turn raw disease facts into operational threat-assessment guidance, making the content more actionable for malicious planning.

Content

No source excerpt is available for this finding.

Harmful Content Injection

Critical
Category
Prompt Injection
Confidence
70% confidence
Finding

This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.

Content

Scanner excerpt · botulism-neurotoxic-food-poisoning/references/botulism-details.md (reported line 6)May include surrounding context.

md
# 肉毒中毒详细参考资料

## 一、毒素特性详解

### 病原学
- **病原体**:肉毒杆菌(Clostridium botulinum)
- **致病血清型**:A型、B型、E型对人致病
- **毒素性质**:剧毒嗜神经外毒素
- **致死量**:对人的致死量仅为0.01mg左右

### 毒素物理化学特性

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill provides concrete diagnostic thresholds, IV/ORS rehydration guidance, antibiotic selection, steroid use, and discharge/isolation criteria for potentially life-threatening infectious diseases without an explicit safety boundary stating it is informational only and not a substitute for urgent clinician evaluation. In this context, omission of a warning is dangerous because users may act on dosing, contraindications, or triage guidance directly, delaying emergency care or causing harm through self-treatment or inappropriate treatment recommendations.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This file gives precise cholera IV fluid formulations, replacement volumes, infusion rates, and potassium dosing with no warning that it is for trained clinicians only, no contraindications, and no instruction to verify against current guidelines. In an agent skill context, this can be surfaced as direct actionable treatment advice to unqualified users or applied without patient-specific assessment, creating substantial risk of fluid overload, electrolyte derangement, arrhythmia, or undertreatment in a life-threatening condition.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

This skill gives concrete treatment, steroid dosing, severity classification, and disposition-oriented output requirements without any visible warning that it is reference information rather than a substitute for licensed medical judgment. In a high-stakes infectious disease context, users may over-trust the output and act on outdated, incomplete, or patient-inappropriate guidance, creating risk of patient harm.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

This skill provides concrete diagnosis-adjacent and treatment instructions for suspected or confirmed anthrax, including isolation, steroid indications, local wound management, airway escalation, and antimicrobial selection, but it does not prominently warn that this is high-risk medical guidance requiring qualified clinician oversight. In a conversational agent context, omission of that warning can lead non-clinicians or underqualified users to act on dangerous instructions during a potentially life-threatening infection, increasing risk of delayed emergency care, inappropriate treatment, or misuse of antibiotics and steroids.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

This skill gives concrete antimicrobial selection, dosing-strategy, organ-function adjustment, pregnancy-use, and FMT recommendations, all of which are high-risk clinical decisions, but it does not include safety warnings, scope limitations, or instructions to defer to licensed clinicians and local guidelines. Because the content concerns resistant organisms and individualized treatment, omission of these safeguards materially increases the risk of patient harm, inappropriate antibiotic use, and dangerous self-medication.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

This file gives highly specific antimicrobial selection, dosing, PK/PD targets, renal adjustment, and multidrug-resistant organism treatment guidance without an explicit warning that it is clinician-facing reference material and not a substitute for individualized assessment, local resistance data, allergy review, organ function review, or stewardship oversight. In an agent skill, this creates a real safety risk because the system may present these recommendations as generally applicable instructions, leading to inappropriate treatment, toxicity, undertreatment, or misuse in patients with different indications or contraindications.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This skill provides concrete diagnosis and treatment instructions, including drug regimens and surgical escalation criteria, without any explicit warning that it is not a substitute for clinician judgment or emergency evaluation. In a medical context, omission of such guardrails can lead users to over-rely on the workflow for self-treatment or delayed care, creating meaningful patient safety risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This skill provides detailed diagnosis and treatment workflow for potentially life-threatening infections such as meningococcal meningitis, pertussis, and typhoid, but does not clearly state that it is informational only and not a substitute for urgent in-person medical evaluation. In this context, omission is dangerous because users or downstream agents may over-rely on the content, delaying emergency care for conditions explicitly described as rapidly fatal or requiring urgent intervention.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file is explicitly centered on bioweapon pathogen classification, threat properties, and response categorization, which materially departs from the stated clinical infectious-disease purpose of the skill. This scope mismatch increases the chance the skill can be used for harmful threat assessment rather than patient care, especially because the content is organized around hazard prioritization instead of diagnosis or treatment alone.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly says its output should guide fluid resuscitation, acidosis correction, and potassium replacement, but it does not include a clear warning that this is clinician-facing decision support requiring independent medical verification. In cholera, errors in rehydration rate, potassium replacement, or acid-base management can rapidly cause shock, arrhythmia, neurologic deterioration, or death, so omission of guardrails materially increases patient safety risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill instructs the agent to generate precise weight-based CMV antiviral regimens, including route, dose, and duration, but does not require clinician verification or clearly warn that this is high-risk medical decision support. This is dangerous because CMV antivirals such as ganciclovir, valganciclovir, and foscarnet have narrow safety margins, require renal adjustment and monitoring, and involve especially sensitive populations such as pregnant patients, transplant recipients, and immunocompromised individuals.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

This file provides specific antiviral dosing and duration guidance for ganciclovir and foscarnet without any explicit warning that treatment must be prescribed and dose-adjusted by a qualified clinician based on indication, renal function, weight, immune status, and monitoring results. Because these are high-risk drugs with narrow safety considerations and the text is framed as a clinical reference, users could over-trust the listed doses and apply them without necessary patient-specific adjustment, creating a real risk of serious toxicity or undertreatment.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill gives specific dosing, escalation thresholds, anticoagulation, immunomodulation, and ECMO trigger criteria for severe COVID-19 management without any explicit limitation that it is for licensed clinicians or that real patients require immediate supervision by qualified medical professionals. Because this is high-risk treatment guidance involving prescription drugs and critical-care interventions, an unqualified or unsupervised user could act on it directly and cause serious harm, especially since some recommendations may be context-dependent, outdated, or require bedside judgment and monitoring.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This skill gives detailed, actionable critical-care treatment instructions for septic shock, vasoactive drugs, anticoagulation, osmotic agents, ventilation thresholds, and artificial liver support without any clear warning that it is not a substitute for licensed clinician judgment. In this context, omission of safety boundaries is dangerous because users may over-trust the skill and apply protocol-like guidance to unstable patients, causing severe injury or death if used outside appropriate expertise, monitoring, or updated standards.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

This document gives detailed shock resuscitation and vasoactive drug dosing instructions, including infusion rates, escalation logic, and organ-support interventions, without any warning that it is high-risk clinical reference material requiring qualified physician oversight and current guideline verification. In an agent skill context, this can enable unsafe automation or over-reliance by non-clinicians, increasing the risk of severe patient harm or death if outdated, misapplied, or used outside an ICU setting.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill issues a blanket 'must perform lumbar puncture' directive for all confirmed pulmonary cryptococcosis cases and explicitly states there are no exceptions, without requiring clinician review of contraindications, patient-specific risk factors, or informed consent. In a clinical decision skill, this can promote unsafe overreach into invasive procedure recommendation, potentially causing harm when LP is contraindicated or when the instruction is followed without appropriate escalation and warning.

Content

No source excerpt is available for this finding.