The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines multiple prompt templates that interpolate external user data into instructions for the agent, creating a surface for indirect prompt injection.
Ingestion points: Placeholders such as [THE PRODUCT], [THE PROBLEM], [TOPIC], and [ADDITIONAL DETAILS] in the Prompt Template sections of SKILL.md are designed to receive untrusted user data.
Boundary markers: The templates do not use specific delimiters or "ignore embedded instructions" warnings to help the AI agent distinguish between template instructions and user-supplied data.
Capability inventory: The skill consists entirely of markdown templates and does not include any scripts, subprocess calls, file-system write operations, or network tools, which significantly limits the impact of a potential injection.
Sanitization: The instructions do not provide any validation, filtering, or escaping logic for the content provided to the placeholders.

ugc-yapper