The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing local Python scripts such as clv-calculator.py, roi-calculator.py, revenue-forecaster.py, and budget-optimizer.py via the shell to perform business logic and KPI calculations. These scripts are invoked with arguments interpolated from data processed by the agent.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection because it ingests data from multiple external and local sources.
Ingestion points: Data is collected from external MCPs (Google Analytics, Google Ads, Meta Ads, HubSpot, Salesforce), campaign execution logs, local brand profiles in ~/.claude-marketing/brands/, and agency SOPs in ~/.claude-marketing/sops/.
Boundary markers: The instructions do not define clear delimiters or specific instructions for the agent to disregard potential commands or malicious formatting embedded within the retrieved marketing data or profiles.
Capability inventory: The agent can execute local shell commands (via Python scripts) and output results to various delivery channels including Slack, Email, and Google Workspace tools.
Sanitization: There is no mention of sanitizing or validating the inputs retrieved from external platforms or local configuration files before they are used in the summary generation process.

exec-summary