The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is primarily designed to fetch and install AI skills from external Git repositories and remote hub JSON files.
Evidence: Found in SKILL.md and references/install.md through commands such as skillshare install user/repo and skillshare hub add https://example.com/hub.json.
[INDIRECT_PROMPT_INJECTION]: The skill manages a pipeline for syncing external rules and instructions into other AI agents (e.g., Claude, Cursor, Windsurf), creating a surface for potential indirect prompt injection attacks from those external sources.
Ingestion points: External Git repositories (install), remote Hub JSON files (hub add), and local project skill sources.
Boundary markers: The tool explicitly includes a built-in audit command and auto-scans after installation to detect malicious content.
Capability inventory: The skill can create symlinks or copy files into active AI agent configuration directories such as ~/.claude/rules and ~/.cursor/rules.
Sanitization: Employs a multi-analyzer audit system described in references/audit.md that scans for destructive commands, obfuscation, and suspicious URLs.
[REMOTE_CODE_EXECUTION]: The tool facilitates the installation and updates of third-party content that may be executed or processed by other AI CLI tools.
Evidence: The install and update commands download content that is then linked into active tool paths.
[COMMAND_EXECUTION]: The skill provides a shell script runner and uses CLI commands to interact with the underlying system and Git.
Evidence: scripts/run.sh executes a local skillshare binary, and the tool uses Git for repository operations.

skillshare