multi-model-analysis
Skill749 starsChanged 14 months agoMITView on GitHub
Runs 3 AI models in parallel to independently analyze a problem and propose approaches. Compares all proposals against a rubric and selects the best. Can be invoked standalone or called from any agent (issue-resolver, pr-review, etc.).
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 33
Before running, show the user:
- Distinctive: line 121
If invoked from an agent, the agent records this in its state file and proceeds with the selected approach.
- Distinctive: line 122
If invoked directly, present the result to the user.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: Multi-Model Analysis
Analyze a problem or code change using three AI models in parallel, then compare and select the best approach.
When to Use This Skill
- ✅ Validate a proposed fix before implementing
- ✅ Review a PR's code changes from multiple perspectives
- ✅ Get a second opinion on an architectural approach
- ✅ Called from another agent to enrich its analysis
Inputs Required
The caller (agent or user) must provide:
| Input | Description |
| Subject | What is being analyzed — a proposed fix, a PR diff, a design approach, etc. |
| Context | Issue/PR description, root cause (if known), affected files and relevant source snippets |
| Primary approach | The default model's or caller's current proposed approach (optional — omit if no prior proposal exists) |
Step 1: Permission Prompt
Before running, show the user:
🤖 Multi-Model Analysis available.
Subject: [one-line summary of what will be analyzed]
Three AI models will independently analyze this and propose approaches.
All proposals will be compared and the best selected.
Proceed with Multi-Model Analysis? (yes / no)- NO → Return to the caller with the primary approach unchanged.
- YES → Continue below.
Step 2: Launch 3 Parallel Agents
Launch all three agents simultaneously using the explore agent type with this prompt:
You are a senior .NET MAUI engineer reviewing code for the Syncfusion Toolkit for .NET MAUI.
Subject: [what is being analyzed]
Context:
[issue/PR description, root cause, affected files]
Source code context:
[relevant source snippets]
[If a primary approach exists:]
Primary approach already proposed:
[paste the primary approach]
Task: Propose your own independent approach. Include:
1. Exact files and methods involved
2. The proposed change (before/after or pseudocode)
3. Why this approach is correct and safe
4. Any risks or edge cases to watch for
[If primary approach exists: 5. Whether you agree or disagree with the primary approach, and why]
Be concise and specific. Do NOT implement — only propose.| Agent | Model |
| Agent A | claude-sonnet-4.6 |
| Agent B | gpt-5.1 |
| Agent C | gemini-3-pro-preview |
Step 3: Evaluate and Select the Best Approach
Compare all proposals (primary + 3 agents) using this rubric:
| Criterion | Weight |
| Correctness — directly addresses the subject | High |
| Minimal change — fewest lines/files affected | High |
| Safety — no regressions, handles edge cases | High |
| Consistency with existing code style | Medium |
| Testability — easy to verify | Medium |
Select the best approach — one proposal or a synthesis of multiple.
Step 4: Report the Result
Return a structured summary to the caller:
## Multi-Model Analysis Result
| Model | Approach Summary | Selected? |
|-------|-----------------|-----------|
| Primary / Default | [brief summary or "N/A"] | ✅ / ❌ |
| claude-sonnet-4.6 | [brief summary] | ✅ / ❌ |
| gpt-5.1 | [brief summary] | ✅ / ❌ |
| gemini-3-pro-preview | [brief summary] | ✅ / ❌ |
**Selected Approach**: [model name, or "synthesis of X+Y"]
**Reason**: [why this approach was chosen]
**Next Step**: [what the caller should do with this result]If invoked from an agent, the agent records this in its state file and proceeds with the selected approach. If invoked directly, present the result to the user.
In this repository2 other files it gives its agents.
- .github/copilot-instructions.mdCopilot instructions
- testing-guideSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.