The Agent Skills Directory

[COMMAND_EXECUTION]: The script invokes the qianwen CLI tool and the macOS security utility using subprocess.run. It correctly passes arguments as a list to the underlying system calls, which prevents shell command injection vulnerabilities that occur when raw strings are interpreted by a shell environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data, specifically support ticket content and engineer replies, through scripts/qianwen_support.py. While the skill does not use specific boundary delimiters for data interpolation, it incorporates strong process-level mitigations: it mandates explicit user confirmation for all state-changing actions, provides instructions for verbatim relay to prevent the agent from executing instructions embedded in ticket messages, and requires masking of sensitive information. The available capabilities include managed command execution and network operations to the support platform.
[EXTERNAL_DOWNLOADS]: The skill connects to the official endpoint cli.qianwenai.com to manage support tickets. These network operations facilitate communication between the agent and the support platform's infrastructure and are necessary for the skill's primary functionality.
[CREDENTIALS_UNSAFE]: The skill manages authentication tokens using environment variables and the system keychain. It implements safety protocols by instructing the agent to mask sensitive data and prohibiting the output of raw credential values, following established security best practices for credential management.

alibabacloud-qianwenai-support