The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a local Python script scripts/validate-views.py intended for cross-view consistency checking. Analysis of the script shows it only uses standard Python libraries (argparse, re, sys, collections, dataclasses, pathlib) and performs read-only operations on markdown files to extract and compare tokens using regular expressions. It does not perform any network operations or unauthorized file system modifications.
[DATA_EXFILTRATION]: No network-enabled tools or commands (such as curl or wget) are present in the skill instructions or scripts. The skill relies on local templates and reference files to generate text-based architectural documentation and Miro prompts for the user to copy-paste manually.
[PROMPT_INJECTION]: The skill's instructions are strictly focused on architectural modeling guidelines. There are no attempts to override safety filters, extract system prompts, or bypass ethical guidelines. It includes explicit instructions to mark assumptions and concerns clearly, which enhances transparency.
[DYNAMIC_EXECUTION]: There is no evidence of dynamic code execution (e.g., eval, exec) or unsafe deserialization. The provided Python script is static and its logic is entirely transparent, focusing on text analysis.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied system descriptions and architectural diagrams, it lacks the dangerous capabilities required to facilitate an exploit. The output is purely informational text and diagrams, and the included validation script does not execute the content of the files it reads.

architecture-models-4plus1