The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to directly execute shell commands provided by the user (short-circuit rule) or generated by the agent, including skills add, npx skills add, and skills remove.
[COMMAND_EXECUTION]: Includes a PowerShell code block intended for execution to verify the success of uninstallation operations by processing JSON output from the skills CLI tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data which could potentially contain malicious instructions.
Ingestion points: Reads output from skills list --json and parses a configuration file src/platforms.ts from a separate installed skill (sync-local-global-agents-skills).
Boundary markers: Implements a "short-circuit rule" that requires verification of command integrity, user intent, and trusted source URLs before execution.
Capability inventory: Includes execution of CLI tools, PowerShell script execution, and file system operations such as creating, backing up, or replacing symlinks and junctions.
Sanitization: Employs cross-referencing against a DEFAULT_PLATFORMS list and manual verification steps to ensure targets are legitimate agent skill directories.

install-skills