Guide · Agent skill
How to use the Academic Claim Verifier agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
academic-verify is an agent skill that checks a research claim or academic citation the way a careful reader would: it looks for the original paper, how the figure was produced, whether the raw data is available, and whether anyone has replicated or challenged it. This guide is for people who have not used agent skills before. By the end you will know what the skill does, how to install it, how to ask for it in plain language, and what to check before trusting the result.
Academic Claim Verifiergarrytan/gbrain · MIT Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from the page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities and upload the skill under Skills.
- 3Install the dependencies. Make sure the perplexity-research skill and the gbrain brain tool are also available, since the skill needs both.
- 4State the claim precisely. Give the exact quote, the source, the number and the time period the claim refers to.
- 5Ask in plain language. Use a trigger such as "check this study" or "is this study real" followed by the claim.
- 6Review the brain page. Read the trace table, the verdict and the caveats before relying on the result.
What the skill makes the agent do
When a book, article or speaker cites a study or a number, the skill traces the claim through four links: claim, publication, methodology, raw data, and independent verification. At each step it asks where the number came from, what the baseline is, whether the raw data is available, and whether anyone has independently verified it.
The skill does not do the web search itself. It sends the claim to the perplexity-research skill for the lookup, then writes the findings into a structured brain page under concepts/ with a verdict: Verified, Partially verified, Unverifiable, Misattributed or Retracted / disputed. The page is a stored note, so later work can reuse the same analysis.
- Claim: the exact quote and who said it
- Publication: title, authors, year and DOI where available
- Methodology: a short summary, with obvious limits flagged
- Raw data: public, proprietary or available on request
- Replication and critique: other studies, retraction or post-publication review hits
Install the academic-verify skill
Download the ZIP for the skill from the page, then unzip it into one of two locations. For a personal install available across your projects, use ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder.
If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
The skill also depends on perplexity-research, which must be installed alongside it. The skill also needs the gbrain brain tool, because its first lookup uses the gbrain query command.
- Option A: unzip into ~/.claude/skills/ for all projects
- Option B: unzip into a project's .claude/skills/ folder for one project
- Option C: upload through Settings → Capabilities → Skills in the Claude apps
gbrain query "<paper title> OR <author name> OR <claim keywords>"Trigger the skill with natural requests
You do not need to type a command. Describe what you want in ordinary language. The skill's description covers requests where a source cites a study and you want to confirm it is real, replicated and accurately described. These phrases are listed as triggers in the skill:
- "verify this academic claim"
- "check this study"
- "academic verify"
- "validate citation"
- "is this study real"
- Mentioning Retraction Watch when you want retraction status checked
Examples of good requests
The more precisely you state the claim, the better the trace. The skill first pins down the exact quote, source, number and time period, so include them if you have them. A vague request such as "is this true" leaves the skill without a target.
- "Check this study: the book says a training programme cut relapse by 40%. Here is the passage."
- "Is this study real? An article cites a 2015 survey for a figure on sleep. Verify it."
- "Validate citation: the policy brief attributes this statistic to a government report."
- "Verify this academic claim and save the verdict to my brain."
Read the brain page and the verdict
The result is a page with a title, a one-line verdict, the exact claim, a trace table with a source for each step, the verdict with reasons, caveats, and see-also links. The verdict labels follow fixed standards. Verified requires the paper to exist, the raw data to be public or independently confirmed, and the citing source to describe the claim accurately. Partially verified means the paper is real but the citation oversells it, for example by calling a correlation a cause. Unverifiable means the number cannot be traced, which is not the same as wrong. Misattributed means the paper does not say what the citation claims.
- Verified: accurate, raw data available, replication exists
- Partially verified: real findings, but the citation overstates them
- Unverifiable: not enough evidence to act on; the skill says "we couldn't verify"
- Misattributed: the cited paper does not support the claim
- Retracted / disputed: retraction, serious expression of concern or documented critique
Safety, network and scope notes
The skill does not run scripts. It does use the network, because the web lookup runs through perplexity-research and the underlying search service. Your queries, including the claim text and any brain context, are sent to that service, so avoid pasting confidential material unless you are comfortable with that.
The skill is built for rigor, not takedown. It states that it does not make claims about a problem without evidence, and a verdict is only as good as the search results behind it. Do not rely on its verdict alone for medical or safety decisions, because it may miss retractions or context.
- Runs scripts: no
- Uses network: yes, via perplexity-research and web search
- Writes pages: yes, under concepts/ (the skill is marked as writing pages)
- Not for open-ended topic research; use perplexity-research directly for that
Tips for better verification
Start with the original source, not a secondary article, when you can. Give the skill the exact wording and the number, because the trace checks the figure against the paper. The skill's own reference list of databases includes Retraction Watch, PubPeer, OSF, Semantic Scholar, OpenAlex and Many Labs, which the agent consults through perplexity-research.
Check the caveats section before you cite the result. If a verdict is Unverifiable, it means the evidence was not found, so treat it as an open question rather than a finding.
- Quote the passage exactly as written
- Include the author, year and title if you know them
- Ask the agent to link the new page to existing people/ pages if they exist
- Reuse saved pages: the skill checks the brain first before searching
What people use it for
- A journalist checking whether a quoted study exists and says what the article claims it says
- A student confirming a citation before using it in an essay or report
- A person maintaining a personal knowledge base who wants research claims recorded with a verification status
- A reader checking a policy brief's statistic against the government report it cites
Frequently asked questions
Is the academic-verify skill free and what licence does it use?
The skill is published under the MIT licence, as listed on the page. Check the licence file in the downloaded ZIP for the full terms.
What do I need to install before using academic-verify?
You need the perplexity-research skill, which the skill uses for the web lookup, and the gbrain brain tool with the gbrain query command for the brain-first lookup. The skill also needs web access through the underlying search service.
Where do I put the skill folder?
Unzip it into ~/.claude/skills/ to make it available across projects, or into a project's .claude/skills/ folder to limit it to that project. In the Claude apps, upload it under Settings, then Capabilities, then Skills.
Does academic-verify run scripts on my computer?
No. The skill is listed as not running scripts. It does use the network, because the web lookup goes through perplexity-research and the underlying search service.
How is this different from perplexity-research on its own?
perplexity-research handles open-ended topic research. academic-verify wraps it with a fixed workflow for a specific claim: it pins down the quote, traces the paper, data and replication, and writes a verdict page. Use perplexity-research alone when you are exploring a topic rather than checking a citation.
Why did the skill return Unverifiable instead of a clear answer?
Unverifiable means the number could not be traced to source data, and no replication or independent confirmation was found. The skill treats this as not the same as wrong. Give it more precise details, such as the exact title or DOI, and run it again.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use academic-verify if you need to check whether a quoted study is real, accurately described and replicated, and you want the result saved as a reusable page. Pick something else if you want open-ended topic research, which perplexity-research handles on its own, or if you need a decision on medical or safety matters, since the verdict relies on web search and may miss retractions or context.
Download the skill