The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform search operations using shell utilities like rg (ripgrep) and grep. These commands interpolate user-provided keywords to query local index files and directory structures.
[DYNAMIC_EXECUTION]: The skill includes instructions to run a local JavaScript maintenance script (node ~/.claude/scripts/generate-index.mjs --write) to regenerate the toolkit index when changes occur.
[EXTERNAL_DOWNLOADS]: The documentation and routing tables reference numerous external resources, including marketplace repositories and official plugins from vendors like Anthropic, Vercel, and Expo. These are presented as pointers for the agent to use if they are already present in the user's environment.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input in the form of keywords to drive its routing and search logic. While this is a standard operational surface, it presents a potential for indirect prompt injection if malicious data is embedded in the searched files or if keywords are not properly handled by the underlying shell tools.
Ingestion points: User-supplied search keywords and task descriptions used for domain routing.
Boundary markers: No explicit delimiters are used in the command templates to isolate user keywords from the shell command structure.
Capability inventory: The skill leverages the agent's ability to execute shell commands (rg, find, grep, node) and various specialized slash commands.
Sanitization: The skill relies on the agent's internal tool-use logic to safely handle keyword interpolation into shell commands.

toolkit-router