The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages a local Node.js server and browser processes through shell scripts. It uses built-in Node.js capabilities to execute system commands for process lifecycle management.
Evidence: scripts/start-server.sh and scripts/stop-server.sh use kill and ps to manage the server. scripts/server.cjs uses execFile to launch browser instances.
[EXTERNAL_DOWNLOADS]: The visual companion interface loads a branding image from a remote domain.
Evidence: scripts/server.cjs contains a hardcoded URL for a logo image hosted at primeradiant.com.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates user-supplied requirements into a design document that is later processed by an automated reviewer subagent.
Ingestion points: User input from the dialogue is written to design files in docs/superpowers/specs/, and browser click events are stored in state_dir/events.
Boundary markers: The subagent reviewer prompt (spec-document-reviewer-prompt.md) does not use distinct delimiters or instructions to ignore malicious content within the design document being reviewed.
Capability inventory: The skill possesses command execution capabilities to manage the local companion server.
Sanitization: The skill does not perform escaping or validation of the content within the design document before it is parsed by the subagent.

brainstorming