The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local Python scripts located within the skill's environment variable path ${CLAUDE_PLUGIN_ROOT}/scripts/. These include approval-manager.py, performance-monitor.py, report-generator.py, and execution-tracker.py. These scripts are used for core functionality like data aggregation, report generation, and managing the mandatory approval gate.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external analytics MCP servers (e.g., Google Analytics, Meta Ads) and outputs formatted reports to external channels like Slack, Email, and Google Sheets.
Ingestion points: Data is pulled from multiple connected analytics platforms via the performance-monitor.py script.
Boundary markers: A mandatory 'Execution Gate' is implemented, requiring the user to explicitly approve a full preview of the report and recipients before any data is sent.
Capability inventory: The skill has the ability to write to the local filesystem (logging and archiving report snapshots) and perform network operations via MCP tools (Slack, SendGrid/Email, Google Sheets).
Sanitization: The primary defense is the human-in-the-loop requirement where the user reviews the 'Execution Summary' before the final delivery step.

send-report