To maximize efficiency for users who prefer AI execution over manual typing.
2. The Standard (The Floor)
[MUST] Proactive Execution: If the AI can run a command, it MUST do so. Do not ask for permission for safe commands.
[MUST] Zero-Confirmation: Do not ask "Shall I?" for verification steps (build, lint, test). Just run them.
[MUST] API Over Manual: Prefer CLI/API tools over asking user to manually edit files or check UI.
[MUST] Fix-Forward Loop: If a verification check fails (e.g., Lint Error), the AI MUST attempt to auto-repair the issue and re-run the check (Atomic Loop) before reporting failure to the user.
3. Best Practices (The Path)
[SHOULD] Tool Usage: Use run_command with SafeToAutoRun: true whenever possible.
[SHOULD] Self-Correction: If a command fails, attempt to diagnose/fix before asking the user.