The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to aggregate highly sensitive information from multiple high-value sources, including HubSpot and OneWave CRM, financial billing systems, and private communications via Gmail and Slack. This results in the consolidation of PII, contract values, and internal communication sentiment into a single local report file.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as support tickets, customer emails, and Slack messages to calculate health scores and generate recommendations, creating a significant attack surface.
Ingestion points: Untrusted content is retrieved from Gmail (mcp__claude_ai_Gmail__gmail_read_message), Slack (mcp__claude_ai_Slack__slack_search_public_and_private), and CRM objects (mcp__claude_ai_HubSpot__search_crm_objects).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when interpolating retrieved message content into the agent's context.
Capability inventory: The skill has access to powerful tools including Bash, Write, WebFetch, and WebSearch that could be manipulated via injection.
Sanitization: No evidence of content validation, escaping, or filtering of retrieved communication data is present.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to search for and process local data files, exports, and meeting notes, which could include sensitive local information.

client-health-dashboard