Security audit
Security checks for vulnerabilities and agentic risk
The skill’s Feishu media-sending purpose is clear, but it tells agents to upload local files through raw curl commands using Feishu app credentials, so users should review it before installing.
Install only if you expect the agent to send selected local or remote media to Feishu. Do not paste Feishu app secrets into chat unless your environment has an approved secret-handling process, and require confirmation of the exact recipient and file path before any upload, especially for videos sent through curl.
The activation condition is broad enough to trigger on common requests about sending files or media, which can cause the skill to engage in contexts the user did not specifically intend. In a skill capable of transmitting local or remote content externally, overbroad activation materially raises the chance of accidental data exfiltration or unsafe tool use.
The skill supports sending local files, remote media URLs, and credential-backed API usage, but it does not warn users about external data transfer, file sensitivity, or how credentials are handled. This omission can lead to users unknowingly transmitting sensitive local content or enabling unsafe secret use without informed consent.
This section explicitly instructs external HTTP transmission to Feishu using curl and app credentials to obtain an access token. In the context of an agent skill, directing raw network requests and secret handling outside the normal tool boundary is dangerous because it can expose credentials, bypass safety controls, and transmit local content without adequate policy enforcement.
message action=send channel=feishu media=https://example.com/image.png message="网络图片"
## 4. 发送视频（重要：需要用 exec+curl 方式）
**⚠️ 注意**：OpenClaw 飞书频道的 `filePath` 参数不支持本地视频文件。
需要用以下 exec+curl 方式发送视频：
The skill documentation instructs operators to use shell commands and direct Feishu HTTP API calls with app credentials, bearer tokens, and local file uploads. This expands the skill from simple message sending into arbitrary command execution and credential handling, increasing the risk of secret exposure, unsafe execution, and unintended external transmission of local data.
The documented curl upload command sends a local video file directly to an external service using a bearer token, which is a concrete exfiltration path for local data. Because this occurs via shell and direct HTTP rather than a constrained media-send primitive, it increases the risk of arbitrary file submission, mishandled authorization, and audit gaps.
TOKEN=$(echo $TOKEN_RESP | grep -o '"tenant_access_token":"[^"]*"' | cut -d'"' -f4)
# 2. 上传视频文件（必须用 file_type=stream）
curl -s -X POST "https://open.feishu.cn/open-apis/im/v1/files" \
-H "Authorization: Bearer $TOKEN" \
-F "file_type=stream" \
-F "file_name=视频文件名.mp4" \
文件的名称、描述和正文均以中文固定表述，未说明是否支持其他语言或允许按用户偏好切换。若组织要求避免未经选择即强制特定语言，这种单一语言约束可能构成自然语言策略问题。
The section on voice messages instructs using ffmpeg and ffprobe to transform and inspect local files before sending them. While adjacent to media sending, this adds a separate file-processing/tool-execution capability that is not stated in the manifest description.
The manifest describes sending files and compressed packages, but these lines go further by instructing the agent to create ZIP and TAR.GZ archives via shell commands. Creating archives is a separate local file-manipulation capability rather than merely sending existing media.