Practical overview
When it fits
Use Research Proof when a promising claim needs a defined test and adversarial review before implementation or a stronger conclusion. It is especially relevant when the evaluator, held-out data, costs or rejection conditions need to be made explicit.
Who it is for
- Researchers reviewing claims in papers, benchmarks or evidence syntheses.
- AI-agent builders who want explicit verifier boundaries and rejection criteria before an agent loop acts.
- Teams tracking how evidence and confidence for a claim change over time.
Documented uses
- Turn a vague research or performance claim into a falsifiable statement with a defined baseline and verifier.
- Identify counterexamples, rejection gates and hidden costs before accepting a result.
- Record evidence, certainty and the next test in a proof ledger.
- Structure review of benchmark, causal, clinical, mathematical or design-research claims.
Requirements & limits
The public GitHub repository lists an MIT license. The README documents Claude Code plugin installation and a skills CLI installation route.
- The workflow is a research-verification skill for agents, not a substitute for running the underlying experiment or validating transfer.
- Its structured outputs include claim, verifier boundary, evidence, rejection gates, verdict and proof-ledger decision.
- The README says references inform the method but are not proof of a user’s claim.
- Claude Code has a documented plugin path; the README also provides a general skills installation route.
- The repository describes JavaScript and lists an MIT license.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Research Proof as a Claude Code plugin
Add the Research Proof marketplace and install its plugin using the documented Claude Code commands.
Before you start
- Claude Code
- A PowerShell-compatible terminal, as shown in the repository commands
Steps
Add the marketplace with
claude plugin marketplace add tonyblu331/research-proof.Install the plugin with
claude plugin install research-proof-plugin@research-proof.Invoke the skill with
/research-proof-plugin:research-proof.
Expected outcome
The documented plugin installation and invocation steps are complete.
Guide limitations
- The documentation provides setup instructions but no demonstrated execution result or compatibility guarantee.
Install the Research Proof skill with the skills CLI
Install the named Research Proof skill using the documented skills CLI command.
Before you start
- Node.js tooling capable of running
npx - An agent skills directory for the manual-install route
Steps
Optionally list available skills with the command below.
npx skills add tonyblu331/research-proof --listInstall the skill with the command below.
npx skills add tonyblu331/research-proof --skill research-proofFor manual installation instead, clone
https://github.com/tonyblu331/research-proof.gitand copyskills/research-proofinto the agent’s skills directory.
Expected outcome
The documented skill installation route is followed; the manual route places the skill directory in an agent skills folder.
Guide limitations
- The README does not specify the skills directory path for individual agents or demonstrate a completed installation.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 4 October 2026
- Original website title
- tonyblu331/research-proof · Official homepage title · retained in its source language