The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/server.cjs script contains logic to execute shell commands via child_process.exec if the BRAINSTORM_OPEN_CMD environment variable is set. This allows for arbitrary command execution if the environment configuration is manipulated by an attacker.
[EXTERNAL_DOWNLOADS]: The Visual Companion server fetches a branding image from https://primeradiant.com/brand/superpowers-visual-brainstorming-logo.png. This request transmits the tool's version string, functioning as a persistent telemetry beacon to an external vendor domain whenever the visual feature is used.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted project data (files, docs, git logs) and renders it in a browser-based visual companion. The lack of sanitization for the generated HTML mockups allows a malicious project file to potentially execute scripts in the user's browser.
Ingestion points: Project source files, documentation, and commit history are read during the context exploration phase in SKILL.md.
Boundary markers: None provided; the agent is not instructed to use delimiters when processing project data.
Capability inventory: The skill writes HTML files to a served directory (scripts/server.cjs) for browser rendering.
Sanitization: The skill does not implement escaping or filtering for external data interpolated into HTML mockups.

brainstorming