The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bash commands that interpolate user-provided input directly into the command line (e.g., --cookie "用户提供的cookie"). This pattern is vulnerable to command injection if the input contains shell-sensitive characters like backticks, semicolons, or dollar signs.
[CREDENTIALS_UNSAFE]: Sensitive session cookies are passed as plain-text command-line arguments to the src.cli module. This practice is insecure as command-line arguments are often visible to other users and processes via system monitors or logged in process histories.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by accepting raw strings from the user and placing them into an executable shell context.
Ingestion points: User-provided strings for the --cookie command-line argument in SKILL.md.
Boundary markers: No boundary markers or 'ignore' instructions are present to delimit the user data from the command structure.
Capability inventory: Use of the bash tool to execute Python CLI scripts in the workspace directory.
Sanitization: The skill lacks instructions for the agent to sanitize or validate the format of the provided cookie before command execution.

xianyu_accounts