The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the git command via subprocess.run to verify source code references associated with diagram components. To prevent command injection, all inputs from the specification files (such as revision hashes and file paths) are strictly validated against hexadecimal patterns and path safety rules in engine/spec.py.
[EXTERNAL_DOWNLOADS]: The fetch_icons.py script downloads official cloud provider icons from the jgraph/drawio GitHub repository. This is a core functionality required to generate vendor-accurate diagrams and targets a well-known, reputable open-source project.
[DATA_EXFILTRATION]: The engine includes a dedicated redact_origin function in engine/evidence.py that automatically strips sensitive credentials from Git remote URLs before they are included in any generated metadata or sidecar files, preventing accidental exposure of repository secrets.
[OBFUSCATION]: The entire codebase is written in clear, documented Python and Markdown. No evidence of hidden instructions, encoded payloads, or deceptive metadata was found during the analysis.

architecture-diagram