Persistent Multi-Perspective Protocol (Roster-Backed, File-Based)

Multi-agent reasoning with adversarial validation is enforced by maintaining a persistent roster of expert personas that govern analysis, synthesis, and risk checks before producing a final output. This protocol requires a Red-Team invalidation attempt and mitigations to prevent shallow responses.

A persistent persona roster is a file-backed ledger of expert personas stored for continuity across turns. It improves decision tracing by mapping recommendations directly to persona input, ensuring tracked tradeoff resolution and explicit agreements in the final action plan.

To resolve tradeoffs and handle second-order effects, apply a protocol that forces first-principles decomposition before solutions and performs tension resolution. This produces explicit agreements, tradeoffs, and an actionable step plan based on multi-perspective synthesis.

Yes, multi-perspective reasoning overhead is calibrated by task scope. The protocol adjusts persona participation intensity for atomic, compound, or systemic tasks, minimizing analysis overhead while still satisfying mandatory decision trace and red-team invalidation rules.

You need a markdown file at .agents/roster.md to maintain persistent memory for multi-agent reasoning. This file-backed ledger stores and updates the stable set of personas, including a Meta Persona, following update-or-when-needed write rules for continuity.

You should not use a roster-backed multi-persona consultation workflow for simple, atomic tasks that do not require tradeoff resolution, second-order effects analysis, or adversarial validation, because the mandatory decision trace and persona synthesis will add unnecessary overhead.