docs: add llm-roundtable to agents (#489)

Co-authored-by: dinhkarate <[email protected]>
Co-authored-by: claude-chan <[email protected]>
This commit is contained in:
authored and GitHub committed 2026-10-17 01:54:43 -04:00
1 parent e6c30b5beb
commit 5ae42a101e
1 file changed
+24
+24
View File
@@ -0,0 +1,24 @@
name: llm-roundtable
repo: https://github.com/dinhkarate/llm-roundtable
tagline: Six models from five vendors vote on choices, plans and PRs when an unattended agent would stop to ask you
description: >-
Agent Skill plus a stdlib-only Python script (MIT). When an unattended coding agent would
stop to ask its user to choose, approve, or review, it convenes a council through each
vendor's CLI: Claude Fable and Opus (claude), GPT (codex), Grok (grok), Gemini (agy), and
the newest free OpenCode model (opencode, no login needed). Round 1 is blind, round 2 is an
anonymous cross-examination (M1..Mn), peers rate each other's evidence 0-2, and a weighted
vote decides with no chairman model; an undecided council falls back to a safe default.
Seats run read-only and sandboxed. Review mode returns APPROVE/CHANGES/REJECT for plans and
PRs. Runs from OpenCode as the host too; --private drops free-model seats that may train on prompts.
tags:
- skill
- multi-model
- council
- voting
- code-review
homepage: https://dinhkarate.github.io/llm-roundtable/
installation: |
```bash
npx skills add dinhkarate/llm-roundtable --skill llm-roundtable
npx llm-roundtable doctor --probe
```