@tooled-app

overwatch enforces mandatory communication during agent execution, requiring status updates at task start, progress milestones, sub-agent spawns, errors, and completion. prevents silent failures and keeps humans informed of work in progress.

anti-hallucination-skill provides a self-monitoring protocol for agents to detect and correct factual errors through active verification, confidence calibration, and metacognitive checkpoints. it prioritizes internal reasoning loops over external validation.

low-token provides three graduated modes for reducing output verbosity under token constraints, from standard conciseness to emergency single-word responses, with auto-escalation based on quota depletion signals.