The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates on data ingested from Google Workspace APIs, which introduces a surface for indirect prompt injection attacks.
Ingestion points: The output_analyzer.py script processes JSON output from Gmail messages (subject, body), Drive file metadata (names), and Spreadsheet contents, all of which could contain malicious instructions from an attacker.
Boundary markers: There are no explicit instructions or delimiters in the prompt templates to instruct the agent to ignore embedded commands within the processed data.
Capability inventory: The skill provides tools to send emails (gws gmail send), modify Drive permissions (gws drive permissions create), and write data to Sheets, which could be abused if an injected instruction is obeyed.
Sanitization: No sanitization or validation logic for external content is mentioned in the provided documentation or script descriptions.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the gws CLI from established sources.
Fetches the package from Anthropic's official NPM registry (@anthropic/gws).
References the official GitHub repository for pre-built binaries (github.com/googleworkspace/cli).
Installs the utility from the crates.io registry using Cargo.
[COMMAND_EXECUTION]: The skill executes multiple local Python scripts for administrative tasks and data analysis.
Runs scripts/workspace_audit.py to evaluate Workspace security settings.
Uses scripts/output_analyzer.py to filter and format command-line output for the agent.
Invokes scripts/gws_doctor.py and scripts/auth_setup_guide.py for environment diagnostics and authentication management.

google-workspace-cli