The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions for the agent to download and execute setup scripts from the vendor's official domain (bailian.aliyun.com) using shell piping methods, such as curl | bash for Unix environments and irm | iex for Windows environments.
[EXTERNAL_DOWNLOADS]: The skill facilitates the global installation of the bailian-cli package via the standard npm registry to enable the tool's core functionality.
[COMMAND_EXECUTION]: The agent is instructed to execute various shell-based utilities including the bl CLI, npm, node, and the GitHub CLI (gh). These commands are used for session telemetry (Call Attribution), version validation, and interacting with the vendor's issue tracker.
[INDIRECT_PROMPT_INJECTION]: The skill processes external and potentially untrusted data through CLI tool outputs and local file ingestion.
Ingestion points: The skill reads metadata.version from local SKILL.md files, parses command-line output from version checks, and processes local files provided by the user (e.g., images or audio files passed to bl commands).
Boundary markers: The protocol mandates wrapping all tool invocations with specific environment variables and requires explicit user confirmation before executing high-risk operations.
Capability inventory: The skill utilizes shell execution for the bl toolchain, performs network operations via curl and npm, and accesses the file system to manage configuration settings.
Sanitization: The skill provides a detailed workflow for redacting API keys, access tokens, and business-sensitive data from log outputs before they are shared in bug reports.

bailian-protocol