The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from arbitrary public websites to synthesize design tokens, which represents a surface for indirect prompt injection attacks.
Ingestion points: Fetches raw HTML and CSS from user-provided URLs as described in the 'How it works' section of SKILL.md.
Boundary markers: The instructions do not specify the use of clear delimiters or instructions for the LLM to ignore embedded commands within the fetched web content.
Capability inventory: The skill uses the memi CLI tool to perform network requests and write output files like DESIGN.md and DesignSpec JSON.
Sanitization: There is no indication that the extracted raw data is sanitized or filtered for instructions before being sent to the LLM for structured synthesis.
[COMMAND_EXECUTION]: The skill relies on the execution of the memi CLI tool to perform its primary extraction tasks.
Evidence: SKILL.md provides usage examples for memi design-doc and memi extract to fetch and parse external site data.

design-extract