The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides links to installation scripts (cli-install.md) and references to external skills from the inference-sh GitHub repository. These are provided as part of the instructional content for setting up the required CLI environment.
[INDIRECT_PROMPT_INJECTION]: The skill defines prompt templates (e.g., Code Review, Content Writing) that utilize placeholders for external data like [code] or [article text]. While this is standard practice for prompt engineering, it creates a surface where untrusted data could influence model behavior if not properly delimited.
Ingestion points: Placeholders in prompt templates within SKILL.md (e.g., [code], [article text], [language]).
Boundary markers: The templates use descriptive labels but lack formal delimiters (like XML tags) or instructions to ignore embedded commands.
Capability inventory: The skill itself is instructional documentation and possesses no automated code execution or file-write capabilities. It relies on the user to run the provided belt CLI examples.
Sanitization: No automated sanitization is performed on the data placed into these templates.

prompt-engineering