Back to skill

Security audit

Arrhythmia Early Warning Analysis Tool | 心律失常早期预警分析工具

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because its export-only option still permits video uploads, and its automatic account creation and credential storage lack clear user control.

Wait for a working export-only option and clearer privacy/account controls before supplying identifiable facial videos. Confirm who can access reports, how videos and credentials are retained or deleted, and any service charges. The artifacts support review concerns, not a conclusion of malicious intent.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (24)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The visible code does not ingest facial video, extract pulse signals, analyze heart rhythms, or produce cardiac risk assessments. Its only substantive method is an unrelated agent-chat stub, and it does not actually execute an external command. Imported base classes and API services are not supplied, so they could contain additional functionality; this assessment is limited to the visible code.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Generic utilities and HTTP access can legitimately support a video-analysis skill; their presence, and the absence of detection logic in a shared utility file, are not independently sufficient to establish a mismatch. However, this chunk also implements an undeclared account and billing workflow: it resolves or creates a persistent identity, sends it to /sys/phoneLogin with silent registration enabled, stores returned credentials, and supplies payment-skill installation instructions on HTTP 402. Those concrete behaviors extend beyond the declared health-analysis purpose. The supplied chunk does not establish whether other files implement the promised analysis, and it does not show facial-video data being transmitted.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied code's substantive functionality is general user-account persistence rather than cardiac screening. Database initialization and path handling alone would be supporting implementation details, but this chunk specifically implements account and token management without an evidenced connection to the declared medical functionality. This assessment is limited to the supplied chunk; other components could implement the advertised analysis. No network transmission or token exfiltration is shown.

Content

No source excerpt is available for this finding.

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: "arrhythmia-early-warning-analysis"
description: "Based on facial video, identifies abnormal rhythms such as premature beats, atrial fibrillation, tachycardia/bradycardia, assists in early detection of heart health risks. | 心律失常早期预警技能,基于面部视频识别早搏、房颤、心动过速/心动过缓等异常节律,辅助心脏健康风险早发现"
version: "1.0.18"
license: "MIT-0"
---

# 💓 Arrhythmia Early Warning Analysis Tool | 心律失常早期预警分析工具

> **智能健康/识别分析中枢** · 图片/视频智能分析 · 结构化报告 · 历史报告云端查询

---

## 🧭 技�

The analysis workflow lacks a privacy warning about external processing of facial videos and cloud-accessible health reports.

High
Category
Output Handling
Confidence
94% confidence
Finding

The workflow directs automatic processing without a clear privacy notice or consent step for transmitting identifiable facial videos and retaining associated health reports remotely. Although cloud APIs are mentioned, users are not adequately informed of sensitive-data recipients, retention, or report access controls.

Content

No source excerpt is available for this finding.

The request helper attaches service credentials to unrestricted caller-supplied destinations.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The helper accepts arbitrary absolute HTTP or HTTPS URLs and then adds service tokens and API keys without checking the destination. If an attacker can influence a caller's URL, credentials can be sent to an attacker-controlled server; accepting HTTP also permits plaintext exposure, potentially compromising access to health-related accounts.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

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.

The documented export-only mode still performs video analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser promises that --export-env-only will not execute analysis, but the flag is never checked. When input is supplied, execution proceeds into analysis despite that promise, potentially transmitting sensitive facial video when the caller expected a non-processing operation.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · skills/smyx_analysis/scripts/skill.py (reported line 28)May include surrounding context.

python
result_json = JsonUtil.parse(result_json_pure_text, result_json_pure_text)

        result_json_common_ai_response = result_json.get("commonAiResponse") if isinstance(result_json,
                                                                                           dict) else result_json
        if result_json_common_ai_response:
            result_json = result_json_common_ai_response

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · skills/smyx_analysis/scripts/skill.py (reported line 33)May include surrounding context.

python
result_json = result_json_common_ai_response

        result_json_health_ai_response = result_json.get("healthAiResponse") if isinstance(result_json,
                                                                                           dict) else result_json
        if result_json_health_ai_response:
            result_json = result_json_health_ai_response

Local video contents are submitted to an analysis API without disclosure that the file will be uploaded.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The implementation reads the entire local video and passes its contents to the analysis API without an upload disclosure or consent check in this path. Facial video used for cardiac screening combines identifiable imagery with sensitive health context, making unexpected external transmission particularly consequential.

Content

No source excerpt is available for this finding.

The documented export-only option still executes video analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only option is accepted but never used to stop execution. With a video input, this command still invokes analysis, potentially uploading identifiable health-related imagery despite explicitly promising not to analyze it.

Content

No source excerpt is available for this finding.

The service exposes generic remote modification and deletion capabilities unrelated to facial-video arrhythmia analysis.

Medium
Category
Excessive Agency
Confidence
86% confidence
Finding

At L077-L081 and L091-L095, the service forwards caller-supplied URLs and arguments to HTTP PUT and DELETE operations. Network access can reasonably support remote video analysis, but general-purpose remote modification and deletion are not justified by the manifest's stated diagnostic purpose.

Content

No source excerpt is available for this finding.

ai_chat documents an agent invocation and analysis result, but never invokes the agent or produces a result.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The docstring at L25–L34 says this method invokes openclaw agent through subprocess and returns an analysis result or session ID. However, the command and subprocess invocation are commented out; L68 initializes an empty dictionary, and accessing result.stderr at L70 raises an exception that is caught and suppressed. The method therefore returns None without performing the documented analysis delegation.

Content

No source excerpt is available for this finding.

FileUtil.open silently truncates existing files without disclosing its destructive behavior.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The generic open helper always uses write mode, which immediately truncates an existing file. No confirmation, user-facing log, docstring, or comment explains this destructive behavior; the helper's name does not distinguish it from a read-only open.

Content

No source excerpt is available for this finding.

Analysis requests can silently register a remote account and persist its credentials, behavior not disclosed by the manifest.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

A normal request can initiate silent remote registration and save returned authentication tokens through UserDao without a separate consent step. This undisclosed identity provisioning and credential persistence is particularly sensitive in a health-analysis workflow, although the snippet does not prove that facial video is sent during registration.

Content

No source excerpt is available for this finding.

Ordinary HTTP requests can silently trigger remote account registration.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The request helper submits 'silent': 1 and 'register': 1 to the login endpoint when credentials are missing, making account creation a hidden side effect of ordinary requests. In a medical-context skill, users should explicitly understand and approve the external identity relationship rather than having it established implicitly.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

This POST sends the resolved identity as both openId and mobile, plus a source identifier, to a remote login/registration service without a visible consent gate. External networking alone is not a vulnerability, but the automatic, undisclosed identity transmission makes this instance a privacy concern; the provided code does not establish transmission of video or medical results at this call.

Content

Scanner excerpt · skills/smyx_common/scripts/util.py (reported line 561)May include surrounding context.

python
"source": ConstantEnum.DEFAULT__SKILL_HUB_NAME
            }
            try:
                _response = requests.post(_url, json=_data)
                if _response.status_code == 200:
                    _response_json = _response.json()
                    if _response_json and _response_json.get("success"):

Configuration loading silently creates directories and writes missing configuration files.

Low
Category
Output Handling
Confidence
91% confidence
Finding

YamlUtil.load creates parent directories and writes a YAML file when the requested path is missing, without a prompt, log, comment, or docstring disclosing this side effect. Because BaseEnum.init_subclass invokes this loader (L053 and L058), importing the module can modify the filesystem even when the caller only expects to load configuration.

Content

No source excerpt is available for this finding.

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 243)May include surrounding context.

python
username = getattr(model, "username", None)
            if username:
                column_names = self.__model__.__table__.columns.keys()
                update_data = {key: getattr(model, key) for key in column_names if key != "username"}
                return self.update_by_username(username, **update_data)

            return None

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 353)May include surrounding context.

python
column_names = self.__model__.__table__.columns.keys()

            for key in column_names:
                value = getattr(model, key)
                setattr(instance, key, value)

            session.commit()

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 326)May include surrounding context.

python
if filters:
                for key, value in filters.items():
                    query = query.filter(getattr(self.__model__, key) == value)

            if offset:
                query = query.offset(offset)

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/dao.py (reported line 441)May include surrounding context.

python
if filters:
                for key, value in filters.items():
                    query = query.filter(getattr(self.__model__, key) == value)

            return query.scalar()
        finally:

Dynamic attribute access via getattr()

Low
Category
Dangerous Code Execution
Confidence
50% confidence
Finding

Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.

Content

Scanner excerpt · skills/smyx_common/scripts/util.py (reported line 30)May include surrounding context.

python
elif isinstance(cur, dict):
                        cur = cur.get(part, default)
                    else:
                        cur = getattr(cur, part)
                return cur
            except Exception:
                return default