The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local Python utility scripts: ../ui-design/scripts/class_rules.py and workflow/scripts/record.py. These scripts are used for analyzing project-specific UI classes and updating the workflow plan. The execution uses the isolated mode flag (-I) to ensure the scripts run in a controlled environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading the project's source code to trace journey paths, failures, and compatibility. This creates an indirect prompt injection surface. The skill mitigates this by adhering to a specific evidence-recording policy in workflow/GROUNDING.md and requiring multi-stage human approval gates before any code modifications occur.

workflow-clarify