What problem does it solve? Organizations running AI agents face unclear ownership, climbing token costs, prototype sprawl, and vendor renewals negotiated without leverage. This Skill provides an operating model for managing agents like managed labor rather than shelfware. ## Core Features & Use Cases - Ownership and maintenance: Assign one accountable owner per consequential agent and run a seven-surface maintenance loop (job, diet, memory, tools, reach, proof, value) ending in an explicit keep/change/pause/retire decision. - Cost and licensing discipline: Measure token usage in three fidelity lanes, read budget burn as signal, score vendor licenses against nine fairness traits, and compute cost per completed unit of work. - Prototype lifecycle management: Classify tools on a four-state ladder (personal, team beta, supported, customer-facing) with promotion and demotion triggers plus a scheduled demotion audit. - Use Case: Before a major AI vendor renewal, inventory all agents, assign ownership cards, run maintenance passes with replay packs, and renegotiate the contract using cost-per-resolved-case data. ## Quick Start Inventory every agent touching real data in my organization, assign an owner to each, and run the first maintenance pass on the highest-cost agent.
