Back to skill

Security audit

Smart Feeder Pet Detection & Recognition Skill | 智能喂食器宠物检测识别技能

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with high confidence because scanners report undeclared account and credential handling, unexpected remote and file-system side effects, and a mismatch between its pet-detection description and some reported capabilities.

Consider the skill suspicious before installing. The reported concerns include remote processing of local video, possible account creation and credential handling, and behavior that may exceed the described pet-identification purpose. Skillspector reports partial analysis because a reference could not be resolved, so the evidence is incomplete; the AIG scan completed with no findings, but that does not resolve the reported concerns.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (25)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

Database persistence could legitimately support the declared pet-registration feature. However, this chunk specifically implements human user-account and token management, a materially different capability not disclosed in the description. Its concrete model and specialized queries concern users, not pets. No camera processing, animal recognition, or pet-identity model is present in the supplied chunk. This assessment is limited to the supplied code; it does not establish what other skill components implement or indicate malicious behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The visible code implements a video-analysis and report-management client. It contains no explicit cat/dog detection, target-area handling, individual pet identification, or pet database enrollment. Local-file reading and API submission could support the declared vision task and are not independently mismatches; however, the report-centric workflow and generic/health response handling indicate a materially different scope. The imported API implementation is not supplied, so pet recognition inside the remote service cannot be ruled out, but it is not established by this chunk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

Generic formatting, polling, filesystem, and HTTP helpers could legitimately support a pet-recognition skill, so their presence alone is not a mismatch. The material discrepancy is the additional account-management behavior: the code can derive an identity from a workspace key file, generate and persist a default user, transmit that identity for silent registration/login, and store returned credentials. It also introduces a paid-service recharge workflow. These behaviors extend beyond the described pet identity database, which concerns animals rather than service-user accounts. The absence of vision code in a shared utility module is not, by itself, evidence that the complete skill lacks the declared capability.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
82% confidence
Finding

The visible implementation is a video-analysis client, with operations for submitting videos and listing historical analyses. It exposes no pet detection, individual pet identification, camera target-area configuration, or pet database enrollment operations. The delegated skill implementation is not supplied, so it could perform pet-related analysis internally; nevertheless, the supplied chunk does not substantiate the declared specialized purpose and instead exposes a materially broader video-analysis workflow. Network requests and optional output-file writing are supporting implementation details, not independent mismatches.

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-detection-feeder-analysis"
description: "Based on computer vision, automatically detects and recognizes cats and dogs appearing in the target area from the perspective of feeder/IPC cameras, and supports pet identity recognition and database entry, suitable for pet identification management in smart feeding scenarios. | 智能喂食器宠物检测识别技能,基于计算机视觉从喂食器/IPC摄像头视角自动检测识别目标区域出现的猫、狗宠物,并支持宠物身份识别和底库录入,适用于智能喂养场景的宠物识别管理"
version: "1.0.26"
license: "MIT-0"
---

# 🍖 Smart Feeder

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 media-analysis workflow lacks a clear privacy warning about remote processing of camera media.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

Although API usage is mentioned, the workflow does not clearly tell users that local camera media is transmitted to a remote processor before analysis. Feeder and IPC footage can reveal household interiors and people, making undisclosed remote processing a meaningful privacy risk.

Content

No source excerpt is available for this finding.

The --export-env-only option promises not to run analysis, but execution proceeds to analysis or enrollment.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The parser accepts --export-env-only and promises no analysis, but execution never checks the flag. When media arguments are supplied, a caller relying on that promise can unintentionally upload private footage or perform persistent pet enrollment.

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 disclosing the upload.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The implementation reads the entire local media file and passes its contents to the analysis service without a visible disclosure or authorization gate in this path. Camera footage may include sensitive household information beyond the pets the user intended to analyze.

Content

No source excerpt is available for this finding.

--export-env-only promises not to analyze video, but does not prevent analysis.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The accepted --export-env-only flag is never used to stop execution despite promising that analysis will not run. With an input supplied, this can transmit private video and initiate remote processing when the caller expects a non-analysis operation.

Content

No source excerpt is available for this finding.

The database layer implements human-user profile and credential management beyond the declared pet-identification purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The User model stores human identity and demographic information, including real names, email addresses, birthdays, and authentication tokens (L448-L465), and UserDao exposes CRUD operations for these records (L470-L473). DAO initialization also creates and migrates the sys_user table (L185-L214), making this an implemented account-management capability rather than pet identity database entry; the manifest does not describe human-account or credential management.

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

Content

No source excerpt is available for this finding.

An ordinary API request can silently register a remote user account, beyond the declared pet-recognition and pet-database operations.

Medium
Category
Excessive Agency
Confidence
92% confidence
Finding

When tokens are unavailable, an ordinary HTTP request can invoke phoneLogin with silent=1 and register=1, creating a remote account without a separate enrollment decision. Account creation is a persistent external side effect not disclosed by the pet-identification description or the generic request interface.

Content

No source excerpt is available for this finding.

The HTTP helper silently submits user identity data to a separate login/registration endpoint.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The generic HTTP helper conditionally sends the selected identity as both openId and mobile to a separate health-service login endpoint, without an explicit caller opt-in. This hidden identity disclosure is not implied by local pet detection or database entry, although the supplied configuration does not establish whether the receiving service is unauthorized.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
93% confidence
Finding

External requests are not inherently unsafe, but this request can transmit API_SECRET_KEY as both openId and mobile because that secret takes precedence when selecting current__user_name. When tokens need initialization, this repurposes a credential as account-profile data at the login endpoint, potentially exposing it through profile storage and logging beyond its intended authentication use.

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"):

Generic history-report phrases can activate this pet-specific skill outside its intended domain.

Low
Category
Trigger Abuse
Confidence
94% confidence
Finding

Generic history-report phrases can trigger mandatory cloud retrieval without establishing that the user means pet reports. This can produce unintended network activity and disclose pet-report information in an unrelated conversation, although it does not by itself demonstrate cross-account access.

Content

No source excerpt is available for this finding.

Loading configuration silently creates directories and writes a configuration file, including during module import.

Low
Category
Output Handling
Confidence
91% confidence
Finding

L017–L020 create directories and write a YAML file when the requested configuration is absent, without a prompt, log, or explanatory comment/docstring. Because BaseEnum.init_subclass calls this loader at L053 and L058, defining the enum subclasses during import can modify the filesystem without disclosing this side effect.

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:

The ai_chat docstring claims it invokes the OpenClaw agent through subprocess, but the implementation never invokes an agent or produces an analysis result.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

L26 explicitly documents a subprocess-based OpenClaw invocation, but the command and subprocess call are commented out at L48-L66. Instead, L68 creates an empty dictionary and L70 accesses its nonexistent stderr attribute; the resulting exception is caught at L81-L82, so the method returns None without performing the documented analysis.

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/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