The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute local Python scripts (plugin-metadata.py and drive-sync-state.py) located within the skill's installation directory (${CLAUDE_PLUGIN_ROOT}/scripts/). These scripts are used to detect the execution environment and manage the tool's configuration state.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the environment and user input to construct a configuration file, which creates a surface where untrusted data could influence the agent's behavior or the tool's settings.
Ingestion points: Discovery of MCP tool names (Step 2), Google Drive folder names and metadata retrieved from the Drive API (Step 3), and user-provided namespace strings (Step 4).
Boundary markers: The instructions do not define specific delimiters or warning instructions for the agent when processing the results of tool discovery or Drive searches.
Capability inventory: The skill has the capability to execute shell commands (via Python scripts), read/write local configuration files, and interact with external cloud storage via Google Drive MCP tools.
Sanitization: The skill instructions do not explicitly describe sanitization or validation steps for the user-provided data interpolated into the JSON blob used for configuration in Step 4.

cowork-setup