The Agent Skills Directory

[COMMAND_EXECUTION]: The module douyin_sign.py utilizes subprocess.run with shell=True to execute commands for the agent-browser CLI. Using a shell to execute commands with string-formatted parameters is a high-risk pattern that can lead to command injection if user-controlled data is not perfectly sanitized.
[INDIRECT_PROMPT_INJECTION]: The skill fetches and displays content from an untrusted external source (Douyin), including user-generated comments and video descriptions, creating a surface for injection attacks.
Ingestion points: Data is retrieved from the DOM or internal APIs in douyin_sign.py and passed to the agent via douyin.py.
Boundary markers: None. External content is interpolated directly into the agent's context without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has the ability to execute shell commands (via agent-browser) and write files to the local system (via the export command).
Sanitization: There is no filtering or sanitization of the retrieved text to prevent malicious instructions from being processed by the agent.
[CREDENTIALS_UNSAFE]: The script douyin_sign.py automatically writes session cookies in plain text to data/douyin_cookie.txt and stores a browser profile in data/profile/. This persistence of sensitive authentication data in the local filesystem increases the risk of credential theft if the environment is compromised.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the user to install agent-browser globally via npm. This is an external, unverified dependency that is required for the skill to function but is not hosted within a trusted organizational registry.

douyin-cli