Back to skill

Security audit

Storage Manager

Security checks for vulnerabilities and agentic risk

Overview

The skill matches its Feishu inventory purpose, but it includes hardcoded Feishu credentials and can read, write, and upload to a preset Feishu table if the user has not configured their own account.

Review carefully before installing. Only use this skill after removing all bundled Feishu credentials and fixed table IDs, rotating the exposed Feishu secret, and verifying that the tool fails closed unless your own FEISHU_APP_ID, FEISHU_APP_SECRET, FEISHU_BITABLE_TOKEN, and FEISHU_TABLE_ID are explicitly configured.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
final_integrated.py:100
Finding

Hardcoded Feishu Credentials Silently Route User Data to a Fixed Tenant

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
install.sh:16
Finding

Unpinned Dependency Installation Creates Supply-Chain Exposure

Content
View full analysis
/dev/null || { echo "安装requests库..." pip3 install requests } ``` The documentation recommends the same unpinned operation: ```bash pip3 install requests ``` ### Technical Analysis The installer does not specify a reviewed version, lock transitive dependencies, require package hashes, or explicitly select a trusted package index. The installed dependency graph can therefore change between installations without a corresponding change to the audited Skill. Although `requests` is a conventional package name and no suspicious custom repository was identified, installation behavior is influenced by the user's pip configuration and package-index settings. A compromised release, compromised transitive dependency, or malicious package served by an untrusted configured index could execute code during installation or later when imported. The use of the standalone `pip3` executable also does not guarantee that the dependency is installed into the same Python environment used to execute the Skill. ### Attack Path 1. The user runs `install.sh` in an environment where `requests` is not already importable. 2. The script invokes `pip3 install requests` without a version constraint or integrity hash. 3. Pip resolves the package and its transitive dependencies using the user's configured index and resolver state. 4. If that source or a resolved release is compromised, attacker-controlled installation or runtime code is installed. 5. The malicious component executes with the privileges of the user running the installer or when the Skill subsequently imports it. ### ...[truncated 521 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (86)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description broadly matches the main product purpose: intelligent location matching, location image handling, and one-click storage into Feishu. However, the declaration lists no permissions, while the code clearly performs significant external operations: it authenticates with Feishu using app credentials, lists all records from a Bitable table, uploads images to Feishu Drive, and creates records in Feishu Bitable. These are materially relevant capabilities and resource accesses that should be declared. The embedded hardcoded credentials also indicate sensitive credential use beyond an empty permission declaration. Therefore this is a description/behavior mismatch driven primarily by undeclared capabilities and resource access, not by an unrelated purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The description broadly matches the core domain behavior: this is indeed a Feishu-based storage manager with intelligent location matching and location image handling. However, there are material gaps. First, the declared permissions are empty, yet the code performs network authentication to Feishu, reads from Bitable, uploads media, and writes records—these are significant external-resource accesses. Second, the description promises '一键入库' (one-click intake), but for new locations the code explicitly stops and returns 'needs_location_photo', requiring a separate upload step before intake can complete, so the behavior is not consistently one-click. Third, the code includes undeclared operational capabilities such as loading all existing records and using embedded/default secrets. Overall, this is a meaningful description-versus-behavior mismatch, though the main product theme is aligned.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The description presents a complete Feishu storage assistant with three major functions: intelligent location matching, location image management, and one-click inventory entry. The supplied code, however, is narrowly focused on the location image management portion only. It loads records from Feishu Bitable, extracts Location and Location_Image fields, caches mappings, and uploads images to Feishu's media API. There is no actual implementation of intelligent matching logic or inventory/record creation for 'one-click storage entry.' Additionally, the code uses network access to Feishu APIs and local filesystem access for reading image files and creating a test file, while the declared permissions are empty. Because the implemented behavior is only a subset of the advertised skill and includes resource access not reflected by the declaration, this is a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

声明描述的是一个更完整的收纳管理技能,至少包含智能位置匹配、图片管理和入库流程;而提供的代码块只覆盖了“智能位置匹配”这一部分,没有图片处理、文件/媒体管理、数据库或表单入库、网络请求、飞书 API 调用等实现。因此代码实际行为明显窄于声明的整体能力,属于描述与实际行为不一致。不过,已实现的那部分“智能位置匹配”与声明中的对应子功能是相符的。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

声明描述的是一个较完整的收纳管理技能,至少应包含智能位置匹配之外的图片管理、入库流程以及可能的飞书侧操作。但提供的代码块只覆盖“智能位置匹配”这一子能力,且是纯算法实现,没有文件/图片处理、网络请求、数据库写入、API 调用、飞书 SDK/接口、触发器或自动化流程。因此描述显著夸大了实际代码能力,属于描述与行为不一致。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The description mostly aligns with the general domain and main functions: intelligent location matching and location image handling for Feishu-based storage records. The code indeed implements fuzzy location matching, checks whether a location has an image, uploads images, and creates records in Feishu Bitable. However, there is a material mismatch in completeness and behavior. The declared title says '完整版' and '一键入库', implying a complete one-click workflow, but for new locations the function explicitly stops and returns 'needs_location_photo', requiring a second submission. Also, location-image management is not durably implemented: add_location_photo stores the mapping only in the in-memory LocationImageManager and does not write a Location_Image field back to Feishu records, so the mapping is lost on restart unless it already existed in the table. Additionally, the declared permissions are empty, but the code authenticates with Feishu, lists Bitable records, creates records, and uploads files to Feishu Drive using credentials, which is undeclared external access. Therefore this should be flagged as a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The description broadly matches the storage-management theme: the code does provide intelligent location matching, image upload/association, and add/update item records in Feishu Bitable. However, the declared permissions are empty while the code clearly authenticates to Feishu, performs remote API calls, reads/writes Bitable records, uploads images, and searches existing records to build location options. These are material capabilities and resource accesses that should be declared. In addition, the code supports retrieval/search and update operations beyond the description’s narrower wording of image management and one-click intake. Therefore this is a description/behavior mismatch, primarily due to undeclared external access and additional data-management capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

描述与核心主题大体相关,确实包含智能位置匹配、图片上传关联和入库功能。但声明明显不完整:1) declared permissions 为空,而代码会访问飞书开放平台认证接口、bitable 记录读写接口、drive 媒体上传接口;2) 代码除“一键入库”外,还提供搜索物品和更新位置两类独立能力;3) 智能位置匹配依赖读取全部现有记录中的 Location 字段,因此具备批量读取表格数据的能力。由于这些都是实际对外能力和资源访问,不只是内部实现细节,因此应判定为描述与行为不完全一致,存在 mismatch。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description presents a complete Feishu storage-management skill with multiple user-facing capabilities: intelligent location matching, image management for locations, and one-click warehousing/storage. However, the actual code chunk is only a test file. It verifies that a FeishuStorageManager can be initialized and can search for a storage item ('梳子') using environment-provided Feishu credentials and Bitable identifiers. There is no visible code for image management, intelligent matching, or one-click入库. The primary purpose of this code chunk is testing configuration loading and search behavior, which is materially different from the declared full-featured assistant. Therefore, this is a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The declared description presents a complete Feishu storage assistant with smart location matching, image management, and one-click inventory entry. The supplied code chunk is instead a standalone test file focused on validating smart location matching and a small search integration. It also directly configures and uses Feishu/Bitable credentials and resource identifiers, which is an externally accessing capability not reflected in the declared permissions. While smart location matching is consistent with part of the description, major claimed features in the description are absent from this code chunk, and the actual primary purpose here is testing/integration rather than delivering the described assistant behavior.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill makes outbound HTTP requests to Feishu APIs, including authentication and record creation, without corresponding declared network permissions. Hidden network capability is dangerous because it enables data exfiltration and remote side effects beyond what users may expect from the skill interface.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill makes outbound HTTP requests to Feishu APIs, including authentication and record creation, without corresponding declared network permissions. Hidden network capability is dangerous because it enables data exfiltration and remote side effects beyond what users may expect from the skill interface.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill makes outbound HTTP requests to Feishu APIs, including authentication and record creation, without corresponding declared network permissions. Hidden network capability is dangerous because it enables data exfiltration and remote side effects beyond what users may expect from the skill interface.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code embeds real-looking Feishu app credentials and tokens as defaults, then uses them for authentication and API access. Hardcoded secrets are a severe vulnerability because anyone with code access can reuse them to access the associated tenant resources, impersonate the application, or extract and modify remote data.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · complete_system.py (reported line 108)May include surrounding context.

python
self.image_manager = None
    
    def get_tenant_access_token(self) -> str:
        """获取Tenant Access Token"""
        if self.access_token:
            return self.access_token

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · final_integrated.py (reported line 116)May include surrounding context.

python
self.image_manager = None
    
    def get_tenant_access_token(self) -> str:
        """获取Tenant Access Token"""
        if self.access_token:
            return self.access_token

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · storage_manager.py (reported line 116)May include surrounding context.

python
self.image_manager = None
    
    def get_tenant_access_token(self) -> str:
        """获取Tenant Access Token"""
        if self.access_token:
            return self.access_token

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · storage_manager_final.py (reported line 154)May include surrounding context.

python
self.image_manager = None
    
    def get_tenant_access_token(self) -> str:
        """获取Tenant Access Token"""
        if self.access_token:
            return self.access_token

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · storage_manager_smart.py (reported line 114)May include surrounding context.

python
self.image_manager = None
    
    def get_tenant_access_token(self) -> str:
        """获取Tenant Access Token"""
        if self.access_token:
            return self.access_token

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code embeds usable default Feishu app credentials and table identifiers directly in source, then automatically uses them if environment variables are absent. Anyone with code access can extract these secrets and authenticate to the linked Feishu resources, potentially reading or modifying records and uploading content to the tenant's storage.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'shell' capability that is not listed in its permissions. This may indicate deceptive intent or missing permission declarations.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · install.sh (reported line 40)May include surrounding context.

sh
CONFIG_EXAMPLE="$SCRIPT_DIR/config.example.env"
cat > "$CONFIG_EXAMPLE" << EOF
# 飞书收纳管家配置
# 复制此文件为 .env 并填写实际值

# 飞书应用配置
FEISHU_APP_ID=cli_xxxxxxxxxxxxxxx

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code embeds Feishu app credentials and table identifiers directly in source as fallback defaults. Anyone with access to the skill can reuse those secrets to access the associated Feishu tenant and bitable, causing unauthorized data access or modification unrelated to the current user.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Using hardcoded Feishu credentials and tokens as implicit defaults means the skill can silently operate against a preconfigured remote account without user awareness. This creates a direct secret exposure and enables unintended access to third-party data stores if the code is reused or executed elsewhere.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The code hardcodes live Feishu app credentials and table identifiers as defaults, meaning anyone with access to the skill source can reuse them to authenticate to the associated Feishu tenant and operate on its data. In a storage-management skill this is not necessary as embedded secrets expose a real external account and can enable unauthorized read/write access far beyond the local tool's function.

Content

No source excerpt is available for this finding.