The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell commands (find, ls) to build a live index of installed components. These commands are provided in references/discovery-commands.md and are restricted to read-only metadata discovery within the agent's configuration directories (~/.claude) and the local project workspace.
[DATA_EXPOSURE]: The scanning process accesses the ~/.claude directory, which contains agent-specific configuration and plugin data. This access is necessary for the skill's primary function of ecosystem discovery and is limited to finding markdown files (SKILL.md, *.md) rather than sensitive credentials or system files.
[INDIRECT_PROMPT_INJECTION]: The discovery process involves reading and summarizing the name and description fields from the frontmatter of any discovered skill or agent. This represents an indirect injection surface where a maliciously crafted local file could attempt to influence the advisor's recommendations. However, the risk is inherent to the scanning purpose and the skill includes quality checklists to encourage manual verification of matches.

using-ecosystem