byted-kickart-subtitle-extractor
Fail
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill implements an update check via
scripts/upgrade.pywhich queries a remote service. According to the instructions inSKILL.md, if a new version is detected, the agent is directed to execute aninstall_commandprovided by the remote API response. This allows for the execution of arbitrary commands controlled by the remote endpoint. - [CREDENTIALS_UNSAFE]: The
SKILL.mdfile contains explicit instructions for the agent to requestACCESS_KEY_IDandSECRET_ACCESS_KEY(Volcengine credentials) from the user in plain text during the chat. It then instructs the agent to runexportcommands to set these as environment variables. This practice exposes sensitive credentials in the session history and process environment. - [EXTERNAL_DOWNLOADS]: The skill uses
curlto download video files from arbitrary URLs provided by the user to the local filesystem (/tmp/openclaw/...) for processing. - [COMMAND_EXECUTION]: The skill relies on multiple shell commands and Python script executions to perform its tasks, including environment configuration (
export), directory management (mkdir), file inspection (file,ls), and package installation (pip install).
Recommendations
- AI detected serious security threats
Audit Metadata