Plan Arbiter
85TrustedInstructions onlyDecision-making for multiple conflicting AI execution strategies and technical proposals.
What it does
Plan Arbiter changes how an assistant handles competing suggestions by forcing a structured evaluation process. Instead of simply presenting options to a user, the assistant acts as an independent auditor that breaks down multiple proposals, checks them against the reality of the codebase, and synthesizes them into a single, actionable path.
This is designed for complex development environments where different models or agents might suggest disparate workflows. By enforcing a standardized memo format, it ensures that risks, rejected alternatives, and verification steps are clearly documented before any code implementation begins.
When it earns its place
Reach for it when…
Use this when you have received multiple strategy documents or code implementation plans from different sources and need to reconcile them into a single, high-confidence course of action. It is effective when plans involve complex dependencies or when you want to cherry-pick the strongest logic from different agents to create a superior hybrid strategy.
Look elsewhere when…
Do not use this for straightforward, single-path tasks where there is only one reasonable way to proceed. It is also unsuitable for rapid prototyping where the overhead of a formal cross-review and decision memo would slow down development on simple or low-risk changes.
What to say to use it
You never invoke a skill directly — the assistant picks it up when your request matches what it covers. These are the kinds of things that do that.
- I have two different proposals for this refactor, can you pick the best one?
- Compare these three strategies and give me the most reliable execution plan.
- Which of these architectural plans handles the dependency issue better?
- Merge the best parts of these two plans into one final set of instructions.
Install it
Install for all your projects
Copy the skill directory into your personal skills folder. Claude picks it up on its own the next time your request matches what it does — there is nothing to enable.
git clone --depth 1 --branch main https://github.com/BuilderIO/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/skills/plan-arbiter ~/.claude/skills/plan-arbiterInstall for one project only
Put it in the project's own skills folder instead, so it can be committed and shared with everyone working on that repository.
cp -r /tmp/skills/skills/plan-arbiter .claude/skills/plan-arbiterIf a skill of the same name exists in both places, the personal one wins.
Installing on Claude.ai or through the API instead? The full guide covers every surface.