Resource name and URL
PolyBrain - multi‑model orchestration skill for Hermes Agent (parallel subagents, citation enforcement, source verification)
https://github.com/mosesman831/PolyBrain
Author (GitHub username or name)
Moses Man (@mosesman831)
Brief description of what it does
PolyBrain is a multi‑agent, multi‑model orchestration layer for Hermes Agent that decomposes a single objective into a JSON task plan, runs research/build subtasks in parallel, then synthesizes and verifies a final answer. It enforces citations, drops uncited claims, and runs a dedicated verifier role to check each claim against its sources before returning the result.
Category it fits into (see README for categories)
Skills & Plugins -> Community Skills
Why it's awesome - what makes it worth including?
It upgrades Hermes from simple delegate_task/Kanban subagents to a multi‑model pipeline where each role (orchestrator, researcher, builder, synthesizer, verifier) can use different LLM/providers that are specialised for the task, so you can mix DeepSeek for research, Claude for synthesis, and GPT‑style models for verification in one run.
It bakes in a “high‑trust” workflow: strict citation enforcement, claim‑by‑claim verification, artifact logging, retries, and robust JSON parsing make it feel closer to Perplexity computer‑style research than a typical ad‑hoc agent script.
It’s config‑driven and reproducible: you install it into .hermes/skills/research/polybrain, tweak config.yaml once, and then Hermes can spin up parallel research/build workflows for any prompt just by calling “Use PolyBrain to …” in chat.
Resource name and URL
PolyBrain - multi‑model orchestration skill for Hermes Agent (parallel subagents, citation enforcement, source verification)
https://github.com/mosesman831/PolyBrain
Author (GitHub username or name)
Moses Man (@mosesman831)
Brief description of what it does
PolyBrain is a multi‑agent, multi‑model orchestration layer for Hermes Agent that decomposes a single objective into a JSON task plan, runs research/build subtasks in parallel, then synthesizes and verifies a final answer. It enforces citations, drops uncited claims, and runs a dedicated verifier role to check each claim against its sources before returning the result.
Category it fits into (see README for categories)
Skills & Plugins -> Community Skills
Why it's awesome - what makes it worth including?
It upgrades Hermes from simple delegate_task/Kanban subagents to a multi‑model pipeline where each role (orchestrator, researcher, builder, synthesizer, verifier) can use different LLM/providers that are specialised for the task, so you can mix DeepSeek for research, Claude for synthesis, and GPT‑style models for verification in one run.
It bakes in a “high‑trust” workflow: strict citation enforcement, claim‑by‑claim verification, artifact logging, retries, and robust JSON parsing make it feel closer to Perplexity computer‑style research than a typical ad‑hoc agent script.
It’s config‑driven and reproducible: you install it into .hermes/skills/research/polybrain, tweak config.yaml once, and then Hermes can spin up parallel research/build workflows for any prompt just by calling “Use PolyBrain to …” in chat.