The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to read context from external files, which presents a surface for indirect instructions.
Ingestion points: Reads from .agents/product-marketing.md, .claude/product-marketing.md, or product-marketing-context.md.
Boundary markers: There are no explicit delimiters or warnings to disregard potential commands found within these context files.
Capability inventory: The skill is primarily instructional and does not utilize capabilities for file writing, network exfiltration, or system command execution.
Sanitization: The content from context files is not validated or sanitized before being incorporated into the agent's task logic.
[SAFE]: The skill provides links to well-known and trusted industry resources for statistical analysis and A/B testing.
References include reputable tools and organizations such as Evan Miller's calculator, Optimizely, VWO, and the American Statistical Association. These are documented neutrally as standard resources.

ab-testing