The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the configuration of an external MCP server at https://rube.app/mcp. This is a third-party endpoint not listed as a trusted or well-known service, introducing a dependency on external infrastructure.
[REMOTE_CODE_EXECUTION]: By adding a remote MCP server, the agent delegates tool execution logic to a third-party service, which provides the underlying functionality for the PostHog automation tools.
[DATA_EXFILTRATION]: The skill facilitates the transmission of analytics data and user events to external PostHog instances through the Rube MCP gateway.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests data from external PostHog APIs that could contain malicious instructions.
Ingestion points: SKILL.md identifies POSTHOG_LIST_AND_FILTER_PROJECT_EVENTS, POSTHOG_RETRIEVE_CURRENT_USER_PROFILE, and POSTHOG_RETRIEVE_FEATURE_FLAG_DETAILS as tools used to read external data into the agent context.
Boundary markers: Absent in SKILL.md. No delimiters or "ignore embedded instructions" warnings are provided for processing retrieved data.
Capability inventory: SKILL.md details write capabilities including creating feature flags via POSTHOG_CREATE_FEATURE_FLAGS_FOR_PROJECT and capturing events via POSTHOG_CAPTURE_EVENT.
Sanitization: Absent in SKILL.md. The instructions do not mention validating or sanitizing retrieved event properties or user profile data before they are processed by the agent.

posthog-automation