The Agent Skills Directory

[PROMPT_INJECTION]: The skill body and metadata contain no instructions that attempt to override agent constraints or bypass safety guidelines.
[DATA_EXFILTRATION]: The analysis script scripts/scan-project.mjs is limited to local file system reads for project inspection and lacks any code for network communication or data transmission.
[COMMAND_EXECUTION]: There is no evidence of arbitrary shell command execution, dynamic code evaluation, or privilege escalation within the skill's scripts or instructions.
[EXTERNAL_DOWNLOADS]: The skill does not fetch or execute any remote scripts or dependencies; it is self-contained and points to official Astro documentation for its audit criteria.

astro-audit