The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill directs the agent to fetch content from an external URL (https://inspira-ui.com/docs/llms-full.txt) for implementation details.
Ingestion points: The agent is instructed in SKILL.md to fetch documentation from an external domain to retrieve props tables and code examples.
Boundary markers: Absent; there are no instructions to ignore malicious directives within the fetched content.
Capability inventory: The skill allows bash, read, glob, and grep tools, which could be misused if the fetched documentation contains malicious instructions that the agent follows.
Sanitization: Absent; the agent is not instructed to validate or sanitize the external code before implementation.
Dynamic Execution (LOW): The script scripts/setup-inspira.sh dynamically generates a TypeScript utility file (utils.ts) using a heredoc (cat > ... << 'EOF'). While this is a standard setup pattern for this type of library, it involves writing executable code to the filesystem at runtime.
External Downloads (SAFE): The skill installs standard, well-known UI and animation packages (e.g., three, gsap, motion-v) via established package managers (bun, pnpm, npm). These are recognized dependencies for the stated purpose of the skill.

inspira-ui