peer-superset

Two-agent peer review improves documentation consensus by coordinating independent reviewer agents to generate separate original critiques, which are then merged into a unified superset of findings and revision requests.

The best way to run independent dual critiques is to instruct two agents to review artifacts independently, exchange their original outputs, and then synthesize the results into a single consensus superset under a designated output directory.

Structured peer-review can be applied to sprint specs, runbooks, and deployment plans, generating artifact-driven outputs with clear phase boundaries and standard done markers for any scenario benefiting from dual critique before synthesis.

You generate a consensus superset by having two independent agents produce original reviews, exchange their artifacts, and then combine the separate findings into a single merged set of revision requests.

This dual-critique workflow requires no external dependencies, operating entirely through phase-based prompts and direct mesh communications to produce artifact-based outputs with standard done markers.

A limitation of independent agent synthesis is that it requires clear phase boundaries and strict artifact management to ensure the two original reviews are genuinely independent before the consensus exchange and merging occur.