Back to skill

Security audit

个性化助眠脑波音乐定制系统

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent sleep-audio recommendation skill, but it activates on very broad sleep mentions and directs users to unsecured health-related audio links.

Review before installing. This skill is intended for Chinese sleep-service workflows and may steer broad sleep comments into a structured assessment. Do not treat its audio plan as medical care, and avoid opening the supplied HTTP audio links unless you trust the network and source; HTTPS, explicit user consent, and clearer language-scope documentation would reduce the risk.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
references/audio-library.md:15
Finding

Therapeutic Audio Resources Are Delivered over Unencrypted HTTP

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:219; references/audio-library.md:15-17, 27-29, 39-41, 51-53, 70
Vulnerability Type: Plaintext transport of health-related audio resources
Risk Level: Medium

Vulnerable Code Snippet

SKILL.md:219:

markdown
> URL prefix: `http://hc.com/cusresources/sleepAudio/`

references/audio-library.md:15-17:

markdown
| Mild | 30min | http://hc.com/cusresources/sleepAudio/bw_general_sleep_pre_30min_mil.mp3 |
| Moderate | 30min | http://hc.com/cusresources/sleepAudio/bw_general_sleep_pre_30min_mod.mp3 |
| Severe | 30min | http://hc.com/cusresources/sleepAudio/bw_general_sleep_pre_30min_sev.mp3 |

references/audio-library.md:27-29:

markdown
| Mild | 15min | http://hc.com/cusresources/sleepAudio/bw_soi_sleep_pre_15min_mil.mp3 |
| Moderate | 15min | http://hc.com/cusresources/sleepAudio/bw_soi_sleep_pre_15min_mod.mp3 |
| Severe | 30min | http://hc.com/cusresources/sleepAudio/bw_soi_sleep_pre_30min_sev.mp3 |

references/audio-library.md:39-41:

markdown
| Mild | 30min | http://hc.com/cusresources/sleepAudio/bw_fna_sleep_pre_30min_mil.mp3 |
| Moderate | 30min | http://hc.com/cusresources/sleepAudio/bw_fna_sleep_pre_30min_mod.mp3 |
| Severe | 45min | http://hc.com/cusresources/sleepAudio/bw_fna_sleep_pre_45min_sev.mp3 |

references/audio-library.md:51-53:

markdown
| Mild | 15min | http://hc.com/cusresources/sleepAudio/bw_ema_sleep_pre_15min_mil.mp3 |
| Moderate | 15min | http://hc.com/cusresources/sleepAudio/bw_ema_sleep_pre_15min_mod.mp3 |
| Severe | 30min | http://hc.com/cusresources/sleepAudio/bw_ema_sleep_pre_30min_sev.mp3 |

references/audio-library.md:70:

markdown
> Full URL prefix: `http://hc.com/cusresources/sleepAudio/`

Technical Analysis

The Skill instructs the agent to provide users with health-related audio files over plaintext HTTP. HTTP does not provide transport encryption, server authentication, or response integrity. A network-positi ...[truncated 2616 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace every http:// audio URL with an https:// URL served by a domain controlled and verified by the service operator.
  2. Configure the audio server with a valid TLS certificate, modern TLS versions, and secure cipher suites.
  3. Redirect HTTP requests to HTTPS only as a migration measure; generate HTTPS links directly so the initial request is not exposed to interception.
  4. Enable HTTP Strict Transport Security after confirming that all required resources are available securely.
  5. Reject or avoid redirects to origins outside an explicit allowlist of trusted HTTPS domains.
  6. Validate the final response origin, expected media type, file size, and filename before treating the response as trusted audio.
  7. Consider publishing a signed manifest containing cryptographic hashes for the approved audio files. Clients capable of verification should reject files that do not match the expected digest.
  8. Avoid embedding symptom and severity information directly in publicly visible filenames. Use opaque resource identifiers or short-lived signed URLs to reduce disclosure of inferred health information.
  9. Verify domain ownership and operational control before deployment, and document a process for certificate renewal, content integrity checks, and incident response.
  10. Add an automated repository check that fails validation when plaintext HTTP resource URLs are introduced.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (9)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · SKILL.md (reported line 3)May include surrounding context.

md
---
name: personalized-sleep-brainwave
description: >
  【个性化脑波音频改善方案】专为睡眠服务机构及从业人员设计的智能问诊与音频匹配工具。
  适用于睡眠门诊、健康管理中心、心理咨询机构、康复机构、企业健康管理、
  养老及高端健康服务等场景,帮助从业人员快速为客户完输出专属脑波音频干预方案。

  ▌ 核心技术优势
  本系统音频基于脑神经调控与功能医学双重技术底座,历经23年持续研发迭代,
  是全球首款非白噪音类个性化助眠脑波音频。有别于市面上雨声、海浪、粉噪

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger is intentionally broad enough to activate on routine, low-specificity sleep mentions such as 'sleep not too good' or 'want to sleep better.' In an agent ecosystem, this can cause unjustified takeover of benign conversations and steer users into a quasi-medical workflow and product recommendation flow without clear consent or relevance gating.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The file is written entirely in Chinese and instructs the agent to conduct the workflow and user-facing dialogue in Chinese-style phrasing, but it does not offer the user a language option. Under the policy, forcing a specific language without opt-in is a natural-language policy concern unless the locale restriction is clearly documented and justified.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown reference is entirely written in Chinese and includes user-facing phrasing to present to users, but it does not indicate that Chinese output is optional, user-selected, or limited to a justified Chinese-only deployment context. The policy requires flagging language or locale constraints when a skill implicitly forces a specific language without opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The file is entirely in Chinese and does not indicate any language selection, fallback, or user opt-in behavior. In a multilingual agent environment, this can cause users to receive medical-adjacent guidance and URLs they may not understand, increasing the risk of misinterpretation, incorrect self-assessment, and unsafe use of the recommended sleep audio plan.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The entire reference is written in Chinese and includes user-facing wording at L94-L96, but the file does not indicate that Chinese is optional or that the skill is intentionally restricted to Chinese-speaking users. Under the policy, a language constraint without user opt-in or clear justification can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

This markdown reference appears to require Chinese comprehension throughout, from the title to all instructions and examples. Under the policy for natural-language violations, forcing a specific language without user opt-in can be a locale-policy issue when no justification or alternative is provided.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

This markdown reference contains only Chinese-language instructions and response templates, but does not state that the skill is China/Chinese-specific or that the user has chosen Chinese. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.