Back to skill

Security audit

Pet Soothing Trigger Analysis Skill | 宠物安抚触发分析技能

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report a mismatch between its advertised pet-soothing purpose and its video-analysis, account, and report-management capabilities, along with privacy and execution-boundary concerns.

Consider the suspicious verdict and the reported mismatch, account handling, and footage privacy concerns before installing. The scan is not fully complete: one local reference could not be resolved, and the behavior of an imported implementation could not be verified. The findings support caution, but do not establish that the author intended harm.

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
scripts/pet_calming_trigger_analysis.py:55
Finding

Default identity resolution bypasses the upstream sender identity

Content
View full analysis

Impact

A second user sharing the Agent workspace can receive reports created under the default account by another user. Subsequent analyses can also be attributed to that shared account.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (31)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code primarily exposes video analysis and analysis-history retrieval, rather than the declared automatic pet-soothing workflow. It calls skill.get_output_analysis and skill.get_output_analysis_list, prints results, and optionally writes them to a user-specified file. Video analysis could support pet monitoring, but this chunk contains no visible pet-specific detection, loneliness tracking, soothing triggers, audio playback, or laser-toy control. The imported implementation is not supplied, so its internal behavior cannot be verified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied code implements a video-analysis and report-management workflow rather than an automatic pet-soothing mechanism. It reads local video contents for API submission or passes a video URL, then retrieves and presents analysis results and report history. Although analysis could potentially support pet care, the chunk does not connect results to anxiety, howling, loneliness, or any soothing-device action. The imported service implementations are not supplied, so their internal behavior cannot be verified; the visible workflow nevertheless differs materially from the declared purpose.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The supplied chunk is a generic database and user-account management module, not a pet-soothing automation implementation. Although persistence can be a supporting detail, the visible functionality centers on unrelated user records, including names, email addresses, birthdays, and tokens, with no demonstrated connection to pet monitoring or soothing actions. The assessment is limited to this chunk; it shows no network transmission or token exfiltration.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The visible code is configuration for AI-analysis and health-report services, with no demonstrated connection to pet anxiety detection, sound playback, or laser-toy control. Although analysis could support pet monitoring, neither that relationship nor a pet-related purpose for the health-report export endpoint is established in this chunk. The mismatch concerns the configured functionality; the code does not itself execute analyses or export reports.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

Generic formatting, polling, and workspace helpers are supporting implementation details and are not themselves mismatches. However, automatic remote account enrollment, credential-file access, token persistence, and payment-related behavior introduce substantial account-management capabilities absent from the declared pet-soothing purpose. The supplied chunk contains no pet-state detection or soothing-device activation. Because it is a shared utility module, that absence alone does not establish that the complete skill lacks those features; the mismatch rests on the additional, concrete account and credential behaviors shown. Most of these behaviors occur when the relevant helpers are called, not merely when the module is imported.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The declared primary purpose is automatic intervention, but the supplied code implements an analysis-and-reporting interface requiring an explicit input or history-list request. It contains no visible monitoring loop, anxiety-trigger evaluation, or device-control action. The imported skill implementation is not supplied, so any calming behavior inside that dependency cannot be verified. Identity resolution and optional output-file writing are supporting implementation details, not independently suspicious capabilities.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied chunk exposes analysis and reporting commands but shows no automatic monitoring loop, distress-based activation logic, sound playback, or laser-toy control. Video analysis and reporting could support the declared purpose, but the visible workflow does not implement its central promised action. The referenced Python module is not supplied, so its internal behavior cannot be verified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

No anxiety detection, howling detection, loneliness tracking, sound playback, or laser-toy activation is implemented in the supplied chunk. The commented-out agent command does not execute and should not be treated as an active capability. Imported base classes may provide additional behavior, but their implementations are not supplied, so they cannot establish the advertised functionality.

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: "pet-calming-trigger-analysis"
description: "Automatically triggers soothing mechanisms (playing relaxing sounds, activating laser toys) when pet anxiety, howling, or prolonged loneliness is detected; a smart companion for pet care. | 宠物安抚触发技能,检测到宠物焦虑、嚎叫、长时间孤独时,自动触发安抚机制(播放舒缓音效、开启激光逗宠),智能宠物陪伴好帮手"
version: "1.0.27"
license: "MIT-0"
---

# 🧸 Pet Soothing Trigger Analysis Skill | 宠物安抚触发分析技能
> **智能分析中枢** · 图片/视频智能分析 · 结构化报告 · 历史报告云端查�

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.

Generic history-query phrases can activate the skill outside pet-care requests.

Medium
Category
Trigger Abuse
Confidence
97% confidence
Finding

Generic phrases such as querying history are coupled to mandatory cloud report retrieval without requiring pet-care context. An unrelated user request could therefore cause an unintended external request and bring private pet-monitoring reports into an unrelated conversation.

Content

No source excerpt is available for this finding.

The local surveillance-video workflow lacks a clear external-processing privacy warning.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The local surveillance-video workflow directs analysis through an external API without clearly warning that the video itself leaves the device. Indoor pet footage can expose people, conversations, and household routines, making ambiguous processing consent particularly consequential.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to perform analysis, but the program still runs analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The program parses --export-env-only but never checks it, so invocation with a valid input proceeds to analysis despite explicitly promising not to do so. A user relying on this non-execution option could unintentionally submit private footage for remote processing; identity initialization also occurs unconditionally.

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

The implemented workflow produces video-analysis reports rather than automatically triggering pet-soothing mechanisms.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The manifest promises automatic soothing actions when anxiety, howling, or prolonged loneliness is detected. At L66-L84, the main workflow submits an analysis, polls for completion, and formats a report; the remaining methods handle video input and report retrieval, with no detection-to-action logic or calls to sound playback or laser controls.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to run analysis, but supplying it does not prevent analysis execution.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The --export-env-only flag is parsed but never checked, so providing it alongside --input or --url still invokes video analysis despite explicitly promising not to. This violates the caller's execution boundary and could cause unintended remote processing of private household footage; actual uploads or device activation are not demonstrated in this source.

Content

No source excerpt is available for this finding.

The service exposes username-based user lookup, a capability outside the stated pet-calming purpose.

Medium
Category
Excessive Agency
Confidence
86% confidence
Finding

At L098–L101, get_user_by_username delegates a caller-supplied username to a user-lookup operation. Looking up user accounts is not a direct requirement of detecting pet distress or triggering soothing devices, and the manifest does not declare account-discovery functionality.

Content

No source excerpt is available for this finding.

The skill includes general-purpose management of sensitive user profiles and tokens, beyond pet-calming automation.

Medium
Category
Excessive Agency
Confidence
89% confidence
Finding

The User model at L448-L465 stores real names, email addresses, birthdays, sex, age, and authentication tokens. UserDao at L470-L473 exposes inherited read, update, and delete operations over these records, while Dao initialization also creates and migrates the user table (L184-L215). This broad identity-management capability is not justified by the manifest's stated purpose of detecting pet distress and activating soothing devices.

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
96% confidence
Finding

The generically named open helper uses mode 'w', which immediately erases an existing file's contents. No docstring, comment, confirmation, or user-facing log explains this destructive behavior; exception reporting does not disclose successful truncation.

Content

No source excerpt is available for this finding.

The request helper silently provisions remote user accounts, a capability not justified by pet-calming automation.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

At L550-L561, the helper posts an identity to a separate health-service /sys/phoneLogin endpoint with silent: 1 and register: 1. When authentication tokens are missing, L589-L604 invokes this registration/login flow and persists the returned user and tokens locally. Silent remote account provisioning goes beyond detecting pet distress and controlling soothing devices, and is not disclosed in the manifest.

Content

No source excerpt is available for this finding.

The HTTP helper silently registers or logs in a remote user and persists returned credentials.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

When tokens are missing, a routine HTTP request invokes a separate phoneLogin call with registration enabled, transmitting the current identity to the health service. It then saves returned user information and tokens locally, without a comment, docstring, confirmation, or user-facing log disclosing these additional account and credential-storage side effects.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
80% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

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"):

The skill introduces an undisclosed payment-skill installation and account-recharge workflow.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

On HTTP 402, the helper returns instructions telling the user to install smyx-payment and recharge a skill account. This is an explicit cross-skill payment solicitation rather than the automatic pet-soothing behavior described in the manifest; the manifest does not disclose a paid-account dependency. The code returns these instructions but does not itself install the payment skill or initiate a charge.

Content

No source excerpt is available for this finding.

Loading configuration silently creates directories and files, including during module import.

Low
Category
Output Handling
Confidence
95% confidence
Finding

YamlUtil.load creates parent directories and writes a YAML file when the requested path is absent, without a prompt, log, docstring, or explanatory comment disclosing this side effect. BaseEnum.init_subclass invokes this loader during class definition (L053 and L058), so importing this configuration module can modify the filesystem unexpectedly.

Content

No source excerpt is available for this finding.