The Agent Skills Directory

[PROMPT_INJECTION]: The instructions and workflow guidelines focus entirely on technical codification of heuristics. No patterns were found that attempt to bypass AI safety filters, override core instructions, or extract system prompts.
[DATA_EXFILTRATION]: The skill does not perform any network operations or access sensitive directories (e.g., .ssh, .aws). All data processing is restricted to the local workspace and project files.
[COMMAND_EXECUTION]: The scripts/test_skill.py file uses subprocess.run to execute the local scripts/classify_conversion.py script as part of an internal test suite. This execution is confined to local files with static test inputs and follows safe development practices.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute scripts from external sources. All logic is contained within the local Python scripts which rely exclusively on the Python standard library.
[SAFE]: The scripts/validate.py utility uses ast.parse for syntax validation, which safely analyzes code structure without executing it. The scripts/classify_conversion.py helper uses regex-based tokenization to score text, presenting no risk of command injection from user input.

heuristic-to-deterministic