The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Bash, Glob, Grep, and Read tools to inspect local files and directory structures to verify plugin compliance. These tools are used appropriately for static file analysis and reporting.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including plugin.json, SKILL.md, and agent Markdown files. While this creates a vulnerability surface, no explicit boundary markers or sanitization logic is provided. The capabilities involved (file reading and bash execution) are restricted to the local filesystem for validation purposes. The identified ingestion points include globbed paths for skills, commands, and agents, while the capability inventory includes Bash, Read, Glob, and Grep.

validate-plugin