What problem does it solve?
Autonomous agent loops that hit their daily token budget normally exit silently, leaving critical work like failing CI or security vulnerabilities unfinished. This Skill lets the agent justify and formally request a budget extension instead of abandoning high-priority tasks.
Core Features & Use Cases
- ROI Calculation: Reads loop-run-log.md to sum actions taken, successful outcomes, and token spend for the day before making any request.
- Structured Escalation: Appends a clearly formatted [BUDGET NEGOTIATION] request to the High Priority section of STATE.md so a human maintainer can make a quick yes/no decision.
- Strict Safety Limits: Caps requests at +20% or +50k tokens, allows only one negotiation per day, forbids self-editing of loop-budget.md, and reverts to report-only mode if declined.
- Use Case: A daily triage loop burns 95k of its 100k token cap but has one critical CI failure remaining; the agent drafts a justification citing its four successful fixes and requests a +50k extension, then suspends until a human approves.
Quick Start
When the loop-budget skill reports 90% or more of the daily token cap is spent and high-priority items remain, activate this skill to draft a budget negotiation request in STATE.md.
