The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing the j-skills CLI tool to manage local files and symbolic links. This provides a direct interface for the agent to modify the local filesystem within project and global configuration directories (e.g., ~/.claude/skills/).
[EXTERNAL_DOWNLOADS]: The documentation instructs users to globally install the j-skills NPM package. It also references the Vercel Labs GitHub repository for path specifications, which is a recognized and well-known service.
[PROMPT_INJECTION]: The skill includes a 'Guidance for LLMs' section. While these instructions are designed to help the agent manage installation scopes (project vs. global), they represent a mechanism for influencing agent decision-making logic.
[INDIRECT_PROMPT_INJECTION]: This skill presents an attack surface for indirect injection as it processes user-provided skill names and paths to execute CLI commands.
Ingestion points: Untrusted data enters via skill names, paths, and environment arguments passed to j-skills install or j-skills link commands.
Boundary markers: The skill documentation does not define specific delimiters or warnings to prevent the agent from misinterpreting embedded instructions within skill files during the 'link' process.
Capability inventory: The CLI tool possesses capabilities to create symbolic links, create directories, and list files across 35+ agent configuration paths.
Sanitization: No explicit sanitization or validation logic is mentioned for the inputs passed from the agent's context to the shell commands.

j-skills