The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The installation instructions in assets/setup.md recommend downloading and executing scripts directly from the service's official domain using curl | bash and irm | iex (PowerShell). While these are standard installation methods for this tool, piping remote scripts to a shell is an execution pattern that carries inherent risk if the source is compromised.
[EXTERNAL_DOWNLOADS]: The skill downloads the bailian-cli package via npm and fetches installation assets from https://bailian.aliyun.com. These downloads target the official infrastructure of the service provider.
[COMMAND_EXECUTION]: The skill extensively uses the bl command-line interface to perform operations such as authentication (bl auth login), configuration (bl config set), and content generation (bl text chat, bl image generate). It also executes system commands like uname -a and node --version to collect environment information for debugging.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content and interpolates it into CLI command arguments (e.g., the --message flag in SKILL.md), which creates a surface for potential instruction injection where malicious data could influence the agent's behavior via the tool's output.
Ingestion points: User-provided text and files are processed in SKILL.md under the 'Respond in the user's language' and 'Local files' sections.
Boundary markers: No explicit delimiters or boundary markers are defined in the instructions for interpolating user data into CLI arguments.
Capability inventory: The skill can execute various CLI commands with network access, file system access, and credential management capabilities.
Sanitization: The skill contains robust redaction guidelines in assets/issue-reporting.md to ensure sensitive data like API keys are removed before being submitted to the external issue tracker.

bailian-protocol