The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various development tools including pnpm, prettier, remark, vale, and tsc via provided bash scripts. These commands are used for validating, formatting, and type-checking documentation and are strictly scoped to the repository's docs/ package or temporary evaluation directories.
[EXTERNAL_DOWNLOADS]: The eval-setup.sh script runs pnpm install to create a sandboxed environment for testing documentation code snippets. This process primarily links local workspace packages from the repository (e.g., @mastra/core) using file-based dependencies, ensuring that the installation is restricted to the user's own project context.
[DATA_EXFILTRATION]: No network exfiltration or unauthorized access to sensitive files was detected. All file operations are confined to the repository worktree and system temporary directories ($TMPDIR or /tmp). The skill uses a RUN_DIR to keep intermediate artifacts outside of the version-controlled repository.
[PROMPT_INJECTION]: The skill instructions do not contain any attempts to bypass safety filters or override agent behavior. The workflow is designed to be interactive and transparent, requiring user approval for the audit scope, findings, and fix plans.

docs-audit