Back to skill

Security audit

unisound-chronic-injury

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly coherent for insurance medical review, but it can send sensitive claim text and a bearer key to any operator-supplied API URL.

Review this carefully before installing in sensitive environments. Only use it with a trusted, intended HTTPS API endpoint, protect and scope the appkey, de-identify claim materials before use, and avoid --output unless you are comfortable storing the question, metadata, and model answer on disk.

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

Error
Location
scripts/run.py:87
Finding

Configurable API Endpoint Can Expose the Bearer Credential and Sensitive Medical Data

Content
View full analysis
str: payload = { "model": model, "temperature": temperature, "messages": [ {"role": "system", "content": system_prompt}, {"role": "user", "content": user_prompt}, ], } try: req = Request( api_url, data=json.dumps(payload, ensure_ascii=False).encode("utf-8"), headers={ "Content-Type": "application/json", "Authorization": f"Bearer {appkey}", }, ) resp = urlopen(req, timeout=timeout) ``` The destination is directly configurable through a command-line argument: ```python p.add_argument("--api-url", default=DEFAULT_API_URL, help="OpenAI 兼容接口地址") ``` The caller-supplied destination and credential are then passed to the request function: ```python out["answer"] = call_llm( api_url=args.api_url, model=args.model, appkey=args.appkey, system_prompt=args.system_prompt, user_prompt=user_prompt, temperature=float(args.temperature), timeout=int(args.timeout), ) ``` ### Technical Analysis The `--api-url` argument accepts an unrestricted URL. `call_llm()` sends both the bearer credential and the complete user prompt to that destination without validating the URL scheme or hostname. There is no enforcement that the destination: - Uses HTTPS. - Matches the documented internal medical-model hostname. - Belongs to an approved endpoint allowlist. - Remains on the approved origin after an HTTP redirect. Consequently, anyone able to influence the command-line ar ...[truncated 2204 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest description and the entire user-facing documentation are written only in Chinese, and the output schema examples require Chinese labels such as “伤情类别” and “判断依据”. This indicates a fixed language/locale behavior without any opt-in, alternative language support, or justification that the skill is restricted to a Chinese-only regulatory or regional context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The script sends full user-provided medical/insurance claim content to an external HTTPS API via urlopen without any built-in consent prompt, masking, or warning that sensitive health data is leaving the local environment. In this skill context, the input commonly contains imaging reports and injury details, so the transmitted data is likely regulated or highly sensitive PHI/PII, making undisclosed remote processing materially risky.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The script writes model outputs and echoed question content to an arbitrary path with write_text/file open, without warning that the file may persist sensitive medical and claim information or overwrite an existing file. Because outputs include question, meta, and answer, the resulting files can store substantial PHI/PII on disk and expose it to other local users, backups, or accidental disclosure.

Content

No source excerpt is available for this finding.