The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting untrusted source code provided through user arguments. * Ingestion points: Target files and modules specified via $ARGUMENTS are read using the Read tool. * Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded prompts in the analyzed code. * Capability inventory: The skill is limited to Read and Grep operations across the skill body and logic. * Sanitization: No explicit sanitization or filtering of the ingested content is performed before processing. The risk is considered negligible as the skill lacks high-impact capabilities like network access or file writing, and the ingestion is central to its functional purpose.
[SAFE]: The skill does not contain any obfuscated code, remote dependencies, or persistence mechanisms. All frontmatter configurations are standard for a code review utility.

naming-obviousness