I built a command-first agent framework where:

if it’s not a command, it’s not allowed

if it’s not listed, it does not exist

ambiguity → abort

No guessing. No fallback. No hidden tools.

This is for people who want agents to behave like deterministic systems, not improvisational assistants.

GitHub: <https://github.com/qiuy-collab/PCAF-Personal-Command-Based-A...>