The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the npx skills command to perform discovery and inspection of external skill packages through search and listing functions.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external, untrusted skill repositories, which establishes an attack surface for indirect prompt injection.
Ingestion points: The agent reads instructions, scripts, and referenced files from external source repositories.
Boundary markers: Explicit instructions are provided to treat all source content as untrusted data and to inventory it for instructions that could override project rules.
Capability inventory: The process involves file system operations (copying and adapting scripts) and command execution via npx.
Sanitization: The skill mandates manual review of all imported code and requires that validation tests be conducted only in isolated temporary workspaces with limited network access.
[DYNAMIC_EXECUTION]: The instructions direct the agent to run scripts from the imported skill to verify behavior, although this is restricted to scripts that have undergone manual review and are executed in isolated environments.
[EXTERNAL_DOWNLOADS]: The skill facilitates the fetching of content from external repositories and uses npx to interact with a skill registry.

kata-import-skill