What problem does it solve?
Multi-agent setups accumulate coordination overhead that often goes unmeasured, and teams keep adding agents without knowing whether each boundary earns its cost. This Skill audits an existing agent topology, counts what each handoff between agents costs, and compares it against what a single capable agent would have produced.
Core Features & Use Cases
- Boundary Cost Analysis: Enumerates every context crossing between agents and names the specific cost at each one, including token burden, handoff loss, compression loss, semantic drift, verification burden, and governance overhead.
- Gain Verification: Checks whether each boundary is earned by a gain a single agent could not produce, such as genuine specialisation, parallelism, cross-vendor diversity, or adversarial review.
- Single-Expert Baseline Comparison: Compares the topology against the null hypothesis of one capable agent doing the whole task, placing the burden of proof on each boundary.
- Use Case: A team running a five-agent review pipeline notices it is slow and agents keep agreeing. The audit reveals two agents whose inputs overlap heavily, recommends collapsing them into one, and identifies which artifact must carry context across the boundaries that remain.
Quick Start
Audit my current multi-agent workflow and tell me whether each agent boundary earns its coordination cost or should be collapsed.
