Security audit
Security checks for vulnerabilities and agentic risk
This video transcription skill needs review because a hostile video link could expose private-network responses, and its network privacy claims overstate the safeguards.
Prefer a version that blocks private-network destinations, validates redirects, and accurately lists upload destinations. Until then, avoid untrusted video links and confidential recordings. Confirm the provider's charges and retention terms, and account for downloaded copies remaining in the local tmp directory.
scripts/utils/download.jsVideo-host redirects can trigger private-network requests and upload their responses
An attacker can induce unauthorized internal GET requests. Readable internal response bodies can also be disclosed to the transcription service through its normal upload flow. Direct delivery of those responses to the attacker is not established.
Token validation can be a legitimate supporting implementation detail, and this utility need not itself perform transcription. However, its private-token activation messaging and promotional referral introduce an access requirement and marketing behavior absent from the declared purpose. The supplied code does not contact the website, transmit the token, authenticate it remotely, or directly terminate execution.
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.
4. 同时传入文件路径与任务ID，优先执行 `--id`，忽略 `--file`
5. 无自定义 prompt 时，默认完整转录视频全部文字
| 用户自然语言指令 | 自动生成命令 |
| -------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| 视频提取 https://example.com/video.mp4 中的文字 | `node scripts/video2text/index.js --file "https://example.com/video.mp4"` |
| 把本地 /path/to/your/video.mp4 改成小红书风格的文案 | `node scripts/video2text/index.js --file "/path/to/your/video.mp4" --prompt "改写成小红书风格的文案"` |
Without declared permissions the skill's intent is opaque and cannot be validated.
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.
当用户用自然语言下达指令时，按以下映射生成命令，保证识别与执行一致：
| 用户自然语言指令 | 生成的命令 |
| ---------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| 提取 https://example.com/video.mp4 里的文字 | `node scripts/video2text/index.js --file "https://example.com/video.mp4"` |
| 总结这个视频的核心观点 https://example.com/video.mp4 | `node scripts/video2text/index.js --file "https://example.com/video.mp4" --prompt "总结这个视频的核心观点"` |
Both removeOnStop and removeOnFail default to true (L0044–L0045), causing the destination to be unlinked on stop here and on failure at L1053. No comment, confirmation, or user-facing message explains this destructive cleanup; it can also delete a pre-existing file supplied to resumeFromFile (L0386–L0388), rather than only a newly created download.