The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/validate_copy.py uses importlib.util to dynamically discover and execute Python modules matching rules_*.py within its own directory. This plugin-style architecture allows for the runtime extension of linting rules based on local code files.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading and analyzing untrusted user content from local files.
Ingestion points: scripts/validate_copy.py and scripts/check_font_coverage.py read the content of files at paths provided by the user or agent during execution.
Boundary markers: The scripts do not use specific delimiters to separate untrusted content from the linter's logic, although mask_code uses regex to skip content inside markdown code blocks.
Capability inventory: The linter scripts can perform directory traversal, local file reading, and local file writing for encoding fixes.
Sanitization: Content is processed via regex matching for specific patterns, and code blocks are masked to prevent them from triggering text-based rules.
[COMMAND_EXECUTION]: The skill instructions guide the agent to run local Python scripts via the shell, passing file paths as arguments to perform validation tasks.

vietnamese-tech-writing