What problem does it solve? Subscription AI plans measure consumption in rolling windows (typically 5-hour and 7-day), but that percentage lives only in the CLI status bar and never reaches the model's context, so the agent plans as if quota were infinite and risks hitting the limit mid-edit with an inconsistent tree and no handoff written. ## Core Features & Use Cases - Quota visibility: A provider-agnostic Python script reads the persisted statusline data (SQLite, read-only) and reports 5h/7d window percentages, reset times, data age, and burn-rate projections in text, one-line, or JSON output. - Tiered planning posture: Defines concrete behavior per usage band (under 60%, 60-80%, 80-90%, over 90%), including when to warn the user, stop opening parallel subagents, and reserve budget for the handoff. - Optional prompt hook: A --gate mode stays silent below a configurable threshold and emits a single warning line when quota crosses it, so the agent is alerted without consuming context in calm sessions. - Use Case: Before starting a long refactor that will spawn parallel subagents, the agent checks the quota, sees the 5h window at 85% with exhaustion projected before reset, writes the handoff while budget remains, and proposes resuming after the reset. ## Quick Start Ask the agent to check how much account quota remains in the 5-hour and 7-day windows and whether the planned task fits before the reset.
