Council is an agent skill by Theo Browne that brings in three strong models from three different providers, usually to review your work, and loops until they're all satisfied.
What it does
Fixes known issues and open PR comments before the review starts.
Runs one flagship model from each provider at high effort, such as Claude Opus, GPT Sol, and Gemini Pro, or the members you name.
Keeps the reviewers read-only and checks each finding against the code as it arrives.
Asks each reviewer for a score out of 10 and runs a fresh round after fixes.
Adds a table of each round's scores to the PR description.
When to use it
Say “council” when a change matters enough to want more than one opinion.
How to get it
The skill is free on GitHub under the MIT license, in Theo's tstack repo. It needs subagents on other providers' models. T3 Code handles that; other harnesses need their own way to run those models.
more skills
Council is an agent skill by Theo Browne that brings in three strong models from three different providers, usually to review your work, and loops until they're all satisfied.
What it does
Fixes known issues and open PR comments before the review starts.
Runs one flagship model from each provider at high effort, such as Claude Opus, GPT Sol, and Gemini Pro, or the members you name.
Keeps the reviewers read-only and checks each finding against the code as it arrives.
Asks each reviewer for a score out of 10 and runs a fresh round after fixes.
Adds a table of each round's scores to the PR description.
When to use it
Say “council” when a change matters enough to want more than one opinion.
How to get it
The skill is free on GitHub under the MIT license, in Theo's tstack repo. It needs subagents on other providers' models. T3 Code handles that; other harnesses need their own way to run those models.
more skills
Council is an agent skill by Theo Browne that brings in three strong models from three different providers, usually to review your work, and loops until they're all satisfied.
What it does
Fixes known issues and open PR comments before the review starts.
Runs one flagship model from each provider at high effort, such as Claude Opus, GPT Sol, and Gemini Pro, or the members you name.
Keeps the reviewers read-only and checks each finding against the code as it arrives.
Asks each reviewer for a score out of 10 and runs a fresh round after fixes.
Adds a table of each round's scores to the PR description.
When to use it
Say “council” when a change matters enough to want more than one opinion.
How to get it
The skill is free on GitHub under the MIT license, in Theo's tstack repo. It needs subagents on other providers' models. T3 Code handles that; other harnesses need their own way to run those models.

