The Agent Skills Directory

[SAFE]: The skill is purely informational and provides guidelines for marketing experiments. No malicious patterns such as prompt injection, obfuscation, or unauthorized data access were found.
[INDIRECT_PROMPT_INJECTION]: The skill includes instructions to read project context from a local file, which is a standard pattern for context-aware agents.
Ingestion points: Reads from .agents/product-marketing-context.md at the start of the workflow.
Boundary markers: No explicit boundary markers are defined within the instruction set to isolate the context data.
Capability inventory: The skill has no high-risk capabilities; it does not perform file writes, shell command execution, or network requests for exfiltration.
Sanitization: Not applicable as the ingested data is used solely to inform the agent's textual response strategy.
[EXTERNAL_DOWNLOADS]: The skill references well-known external tools and services such as PostHog, VWO, and a reputable statistical calculator (evanmiller.org). These references are for user informational purposes and do not involve automated execution of untrusted code.

19-ab-test-setup