The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell scripts and system utilities to perform its tasks. In scripts/make-app.sh, it uses defaults, killall, osacompile, and mdimport. In assets/open.sh, it uses curl, nohup, and open. The instructions also suggest using strings and grep on local binaries to find configuration settings.
[PERSISTENCE]: The skill maintains a presence on the user's system by pinning applications to the macOS Dock and starting background servers that survive session termination. Specifically, scripts/make-app.sh modifies com.apple.dock to add new application tiles, and assets/open.sh uses nohup to ensure the Bun server continues running after the launcher script exits.
[DYNAMIC_EXECUTION]: The skill generates executable application bundles at runtime using system tools. The script scripts/make-app.sh utilizes the osacompile utility to wrap a shell script into a double-clickable macOS .app bundle via AppleScript.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by reading and writing local files which could contain untrusted data that influences the agent's behavior.
Ingestion points: assets/server.ts reads data from local configuration files (e.g., ~/.example/config.json) and provides it to the agent via a GET route.
Boundary markers: None are present in the code; the agent is instructed to read "real files" directly to determine state.
Capability inventory: The skill can write files (writeFileSync in server.ts) and execute shell commands via the generated app launcher and background server.
Sanitization: The server performs basic JSON parsing but lacks rigorous validation against malicious content within the files it manages, relying on the agent to interpret the data correctly.

dock-app