The Agent Skills Directory

[SAFE]: The skill provides legitimate documentation and commands (pnpm, node) for Astro framework development. No security risks were identified across prompt injection, data exfiltration, or remote code execution categories.
[COMMAND_EXECUTION]: The skill provides standard development commands such as pnpm install, pnpm run build, and pnpm run test. These are appropriate for the skill's context as a developer guide for the Astro monorepo and do not represent a security risk.
[INDIRECT_PROMPT_INJECTION]: The evals/evals.json file contains sample prompts for testing the agent's knowledge, representing a data ingestion surface.
Ingestion points: Technical prompts in evals/evals.json (e.g., architecture design scenarios).
Boundary markers: Absent; instructions are processed as plain text.
Capability inventory: Shell execution capabilities provided in SKILL.md (pnpm, node).
Sanitization: Absent; the data is treated as trusted development context.

astro-developer