skills.rest
diazMelgarejo
diazMelgarejoCOMMUNITY
Explore all skills by this author →
Website𝕏@Melgarejo_yDiazGitHub

llm-council-orchestration

Coordinates a three-model LLM council with gated review loops for plan and code validation.

Updated Mar 22, 2026
Search Tags:#code review#delegation#multi-agent#prompt templates#lm studio#llm council#review gates
One-click install
npx skills add https://github.com/diazMelgarejo/orama-system --skill llm-council-orchestration-diazmelgarejo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-council-orchestration
Source: https://github.com/diazMelgarejo/orama-system/tree/main/bin/orama-system/skills/archive/llm-council-orchestration-absorbed
Command: npx skills add https://github.com/diazMelgarejo/orama-system --skill llm-council-orchestration-diazmelgarejo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? This archived Skill documents how to coordinate multiple LLMs (a host executor, a strict reviewer, and a local specialist) so that plans and code deliveries pass structured review gates instead of shipping unverified output. Note: this Skill is superseded and redirects to pt-orama-council and hermes-council-review-gates. ## Core Features & Use Cases - Three-Model Council Roles: Defines Host/Executor (Codex/orama), Reviewer/Critic (Antigravity/Gemini), and Local Specialist (Qwen via Hermes and LM Studio) with ready-to-paste prompt templates. - Gate-Based Review Workflow: Enforces mandatory review gates at initial plan, architecture, core implementation, tests, and final review, looping until the reviewer marks output CLEAN. - Delegation Patterns: Supports delegating private or heavy sub-tasks to a local model, with patterns like Gemini plans/Hermes executes or Hermes drafts/Gemini reviews. - Use Case: A developer planning a risky refactor sends the plan to a reviewer model, iterates until approved, delegates sensitive local file operations to a local Qwen model, and only commits after a final CLEAN verdict. ## Quick Start Read the redirect targets pt-orama-council and hermes-council-review-gates instead of this archived card, since this Skill is superseded and kept only for historical reference.

Frequently Asked Questions about llm-council-orchestration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up a multi-model LLM council for code review?▼

Assign three roles: a Host/Executor that plans and writes code, a Reviewer that critiques using a fixed structure (strengths, critical issues, revisions, approval status, score), and a Local Specialist for delegated sub-tasks. Pass review packages between them at defined gates.

What review gates should an LLM council workflow include?▼

The minimal gate set is initial plan, architecture/design for risky changes, core implementation, tests/verification, and a final pre-commit review. Each gate loops back for revision until the reviewer marks the delivery CLEAN.

Is the llm-council-orchestration skill still active?▼

No, it is superseded. The workflow was absorbed into pt-orama-council and the hermes-council-review-gates reference, and this archived card exists only for historical context and should not be used directly.

Which local model does the council use for delegated tasks?▼

The canonical local coder model is qwen3.5-27b-claude-4.6-opus-reasoning-distilled-v2 running via LM Studio. The archived prompts mention a Qwen 3.6 Coder name, which was corrected as an invented model name.

When should I delegate sub-tasks to a local model instead of the host?▼

Delegate when sub-tasks involve private or sensitive content, or heavy local operations like file writes and terminal commands. The local specialist returns verified output to the host, which remains responsible for the final reviewed delivery.

Related Skills

Explore More Software Engineering Skills→
akillnessakillness
40

lmstudio-cli

Verify LM Studio endpoints and enumerate model IDs from /v1/models.

Community
Advanced
JerrylalalaJerrylalala
6

todo-resolve

Resolves approved todo items in parallel and documents lessons learned.

Community
Advanced
alexiolanalexiolan
17

llm-review

Orchestrates local LLM task lifecycle including load, execute script, collect results, unload via agent dispatch.

Community
Advanced
rachittshahrachittshah
27

council

Coordinate cross-model deliberation across frontier models to produce a trusted answer.

Community
Advanced