ClawHub
Back to skill

Security audit

fangcloud-api skills

Security checks for vulnerabilities and agentic risk

Overview

This Fangcloud skill matches its stated cloud-file purpose, but it gives an agent broad file/admin power, runs an unverified downloaded CLI, and includes real-looking bearer tokens in its API docs.

Install only if you trust the publisher and Fangcloud release host. Use a least-privilege user token, avoid setting an admin token unless you truly need admin APIs, verify or pin the downloaded CLI before running it, and manually confirm any upload, download, share, collection, invite, delete, move, or admin action.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (28)

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The documentation states a file-delete endpoint but shows a curl example targeting a folder deletion URL instead. This can cause consumers or downstream agents to invoke the wrong destructive API and delete the wrong resource type, creating integrity and availability risk for stored content.

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The permanent file-deletion section documents a file trash-deletion API but the example calls the folder trash-deletion URL. Because this operation is irreversible, inconsistent documentation materially increases the risk of accidental permanent deletion of the wrong resource type.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The file-move section shows a folder move URL in the sample request, contradicting the declared file endpoint. This can lead integrators or automated agents to move folders instead of files, causing unintended reorganization or exposure of larger data sets than intended.

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The restore-from-trash example for files points to a folder restore URL rather than the documented file restore path. In a recovery workflow, this mismatch can restore the wrong object type and lead to unintended data resurfacing or failed recovery of the intended file.

Intent-Code Divergence

Medium
Confidence
96% confidence
Finding
The file-update section provides a folder update URL in the curl example instead of the stated file update path. This can cause unauthorized or accidental modification of folder metadata when users intend to update a file, especially in automated tooling that copies examples verbatim.

Missing User Warnings

Low
Confidence
91% confidence
Finding
The skill explicitly describes downloading remote file content and saving it to local disk, but it does not require an explicit user-facing confirmation or warning before writing files locally. This can lead to unintended local persistence of sensitive or untrusted content, especially in an agent setting where users may not realize a cloud operation results in local filesystem changes.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs uploading local files to a cloud service without an explicit privacy, data-transfer, or consent warning. In practice, this can cause accidental exfiltration of sensitive local files if the user does not fully understand that the action moves data from the local environment into a third-party cloud platform.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill includes creating external share links but does not prominently warn about potential data exposure, link scope, or whether the link may be publicly accessible. This creates a real risk of unintended disclosure if a user generates a share link for sensitive files without understanding the access implications.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The documentation presents irreversible deletion behavior without a prominent safety warning, confirmation guidance, or rollback limitations. In agentic or copy-paste usage, this increases the chance that users execute permanent deletion commands without understanding the impact.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The document includes many credentialed API examples, several with bearer tokens embedded directly in sample commands, but provides no token-handling or privacy guidance. This encourages unsafe practices such as hardcoding secrets into docs, prompts, scripts, logs, or chat histories, increasing the risk of credential leakage and unauthorized access.

External Transmission

Medium
Category
Data Exfiltration
Content
| `id` | `path` | `int64` | 文件的ID |
| `target_folder_id` | `body` | `int64` | 目标文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/75003907526/copy' \
Confidence
99% confidence
Finding
This sample embeds what appears to be a live bearer token directly in documentation. Publishing reusable credentials in example commands can enable unauthorized access, data manipulation, and token replay if the secret is valid or copied into derivative systems and logs.

External Transmission

Medium
Category
Data Exfiltration
Content
| `id` | `path` | `int64` | 文件的ID |
| `target_folder_path` | `body` | `string` | 路径名，以/划分文件夹，根目录在个人文件下，文件夹名称必须是1到222个字符，并且不能含有/ ? : \* " > < |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/75003907526/copy_by_path' \
Confidence
99% confidence
Finding
A bearer token is included inline in the sample request, creating a direct secret exposure risk. Even if expired, publishing realistic tokens normalizes insecure behavior and may leak operational credentials if copied from real environments.

External Transmission

Medium
Category
Data Exfiltration
Content
| `parentFolderId` | `body` | `int64` | 父文件夹Id |
| `type` | `body` | `string` | 文件类型；1：doc类型; 2：ppt类型; 3：xls类型 |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/create_blank_file' \
Confidence
99% confidence
Finding
The example contains a bearer token in cleartext within the docs. This is dangerous because documentation is widely replicated, indexed, and ingested by agents, making secret leakage persistent and difficult to contain.

External Transmission

Medium
Category
Data Exfiltration
Content
| `id` | `path` | `int64` | 移动文件的ID |
| `target_folder_id` | `body` | `int64` | 目标文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/179000000087/move' \
Confidence
99% confidence
Finding
This sample includes a cleartext bearer token in a command for a move operation. Exposed credentials combined with write-capable endpoints increase the chance of unauthorized modification or disruption of stored content.

External Transmission

Medium
Category
Data Exfiltration
Content
| `name` | `body` | `string` | 文件名称，文件名称必须是1到222个字符，并且不能含有/ ? : \* " > < \\ |
| `upload_type` | `body` | `string` | 上传类型；固定传api |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/75003685568/new_version_v2' \
Confidence
99% confidence
Finding
A real-looking bearer token is hardcoded in the new-version upload example. Because this endpoint can modify content and returns upload URLs, leaked credentials may permit unauthorized file replacement or staged data upload.

External Transmission

Medium
Category
Data Exfiltration
Content
| --- | --- | --- | --- |
| `item_typed_ids` | `body` | `array<string>` | 文件id或文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/pack_download' \
Confidence
99% confidence
Finding
The pack-download example exposes a bearer token in plaintext. A valid token on a bulk-download endpoint can enable large-scale data exfiltration, making this more dangerous than a simple read-only metadata call.

External Transmission

Medium
Category
Data Exfiltration
Content
| --- | --- | --- | --- |
| `id` | `path` | `int64` | 更新文件的ID |

**Curl Command**:

```bash
curl --location 'https://open.fangcloud.com/api/v2/folder/179000000087/update' \
Confidence
99% confidence
Finding
This update example hardcodes a bearer token in public-facing documentation. Because the endpoint changes resource metadata, leaked credentials may be used for unauthorized edits and can be replayed by anyone with access to the docs.

External Transmission

Medium
Category
Data Exfiltration
Content
| `target_folder_path` | `body` | `string` | 路径名，以/划分文件夹，根目录在个人文件下，文件夹名称必须是1到222个字符，并且不能含有/ ? : \* " &gt; &lt; |
| `upload_type` | `body` | `string` | 上传类型；固定传api |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/upload_by_path' \
Confidence
99% confidence
Finding
The upload-by-path example includes a bearer token directly in the curl command. Credential exposure on upload-related endpoints can permit unauthorized content injection and abuse of returned upload URLs.

External Transmission

Medium
Category
Data Exfiltration
Content
| `parent_id` | `body` | `int64` | 上传至的文件夹id |
| `upload_type` | `body` | `string` | 上传类型；固定传api |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/file/upload_v2' \
Confidence
99% confidence
Finding
A bearer token is present in a file upload example. If valid, it could be used to create or stage unauthorized content uploads, and even expired tokens reveal poor secret-handling practices that may recur elsewhere.

External Transmission

Medium
Category
Data Exfiltration
Content
| `id` | `path` | `int64` | 文件夹id |
| `target_folder_id` | `body` | `int64` | 目标文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/179000000087/copy' \
Confidence
99% confidence
Finding
This folder copy sample publishes a bearer token inline. Since the endpoint can duplicate content into other locations, an exposed token may enable unauthorized replication of sensitive data.

External Transmission

Medium
Category
Data Exfiltration
Content
| `name` | `body` | `string` | 文件夹名，文件夹名称必须是1到222个字符，并且不能含有/ ? : \* " &gt; &lt; \\ |
| `parent_id` | `body` | `int64` | 父文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/create' \
Confidence
99% confidence
Finding
The folder creation example contains a hardcoded bearer token. Leaked credentials can be abused to create unauthorized structures, stage data, and verify account access against the platform.

External Transmission

Medium
Category
Data Exfiltration
Content
| --- | --- | --- | --- |
| `target_folder_path` | `body` | `string` | 路径名，以/划分文件夹，根目录在个人文件下，文件夹名称必须是1到222个字符，并且不能含有/ ? : \* " &gt; &lt; |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/create_by_path' \
Confidence
99% confidence
Finding
A bearer token appears in cleartext in this create-by-path sample. Publishing credentials in markdown exposes them to broad reuse, indexing, and AI ingestion, increasing the chance of unauthorized access.

External Transmission

Medium
Category
Data Exfiltration
Content
| `id` | `path` | `int64` | 文件夹id |
| `target_folder_id` | `body` | `int64` | 目标文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/179000000087/move' \
Confidence
99% confidence
Finding
This move-folder sample includes a bearer token directly in the documentation. Because move operations can alter data organization and access patterns, a leaked token could be used to disrupt workflows or relocate sensitive content.

External Transmission

Medium
Category
Data Exfiltration
Content
| --- | --- | --- | --- |
| `id` | `path` | `int64` | 文件夹id |

**Curl Command**:

```bash
curl --location --request POST 'https://open.fangcloud.com/api/v2/folder/179000000087/update' \
Confidence
99% confidence
Finding
The folder update example exposes a bearer token inline. This can enable unauthorized metadata changes and confirms that secret-handling in the documentation process is unsafe.

External Transmission

Medium
Category
Data Exfiltration
Content
> 请求示例

**Curl Command**:

```bash
curl --location 'https://open.fangcloud.com/api/v2/knowledge/chatStream' \
Confidence
99% confidence
Finding
The knowledge chat example hardcodes a bearer token in the documentation. Since chat requests can include sensitive prompts, library IDs, and session data, an exposed token can enable unauthorized access to AI-backed knowledge resources and data flows.

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
references/openapi.md:348