The Agent Skills Directory

[SAFE]: The skill implements a robust four-phase workflow (Understand, Propose, Generate, Validate) which includes a mandatory confirmation step (Phase 2: Propose) before any files are generated. This human-in-the-loop design prevents the automated execution of potentially malicious or unintended requests.
[SAFE]: The subagent architecture (Generator, Validator, and Fixer agents) is controlled by specific instructional markdown files that limit the agent's scope to structural data formatting (XML and JSON). There is no evidence of arbitrary command execution, network exfiltration, or unauthorized file system access.
[SAFE]: All external references, such as installation commands and repository links, point to the official infrastructure of the author (luongnv89). These follow standard vendor resource patterns and do not involve untrusted third-party dependencies.
[SAFE]: The skill includes comprehensive validation logic (Phase 4) that performs automated quality checks on generated artifacts. This ensures the output files (.drawio and .excalidraw) are well-formed and compliant with the target application schemas, reducing the risk of generating malformed data.

diagram-generator