Guide · Agent skill
How to use the BiasCheck agent skill: install and run it
Updated · written with AI from the project's own documentation and checked against our catalogue data
BiasCheck is an agent skill that audits a web article, document, or pasted text for bias in three layers: the data and study methods, the conflicts of interest of the organization behind the source, and distortions added by journalism. This guide explains what an agent skill is, how to install BiasCheck, how to ask for a check in plain language, and what the skill needs and does on your machine. By the end you should be able to run a bias check and read its output.
BiasCheckdanielmiessler/LifeOS · MIT Download the skill Quick start
- 1Download the ZIP. Download the BiasCheck ZIP from this page and unzip it on your computer.
- 2Place the folder. Move the BiasCheck folder to ~/.claude/skills/ for all projects, or to .claude/skills/ inside a specific project.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill there instead.
- 4Confirm the files are present. Check that SKILL.md, BiasTaxonomy.md, and Intents/Check.md are inside the BiasCheck folder.
- 5Make a request. Ask in plain language, such as "bias check" followed by a URL, file path, or pasted text.
- 6Review the three layers. Read the data, source-organization, and journalism findings, then the supported-versus-editorialized split.
What is an agent skill and what does BiasCheck do?
An agent skill is a folder containing a SKILL.md file that tells a compatible AI agent, such as Claude Code, when and how to perform a specific task. Once installed, the agent loads the skill when your request matches its trigger phrases.
BiasCheck runs a three-layer bias audit on a URL, a file path, or raw text. It fetches the content and any study it cites, then checks biases inside the data, conflicts of interest in the source organization, and distortions the journalism added. The output separates what the data actually supports from what was editorialized, so the original claim can be restated more accurately.
- Layer 1: biases inside the data, such as source or funding bias and sampling bias.
- Layer 2: source-organization biases, including conflicts of interest.
- Layer 3: biases added by the journalism or commentary.
- A reference catalog of bias types is included to name each problem precisely.
Install the BiasCheck skill from the ZIP
Download the BiasCheck ZIP from this page and unzip it. Then place the BiasCheck folder in one of the following locations, depending on whether you want it available everywhere or only in one project.
For all your projects, put it in ~/.claude/skills/. For a single project, put it in that project's .claude/skills/ folder. The skill folder should contain SKILL.md along with its other files (BiasTaxonomy.md and Intents/Check.md).
- Personal install: ~/.claude/skills/BiasCheck/
- Project install: .claude/skills/BiasCheck/ inside your project folder
- Claude apps: upload the skill under Settings → Capabilities → Skills
~/.claude/skills/BiasCheck/Trigger the skill with natural requests
You do not need a special command. Ask in plain language and include the thing to check, either as a link, a file path, or pasted text. The skill's trigger phrases include bias analysis, analyze bias, bias check, check this study, who funded this, is this source biased, fact-check article, methodological flaws, source credibility, and what's wrong with this claim.
The skill's Check intent is triggered by phrases such as "bias check", "analyze bias on", and "check this study/source/article".
- "bias check https://example.com/article"
- "check this study: [paste abstract]"
- "who funded this report? ~/Downloads/report.pdf"
- "is this source biased: [paste text]"
bias check https://news.example.com/some-article-citing-a-vendor-surveyRead the output
The report covers the three layers separately. Each bias claim should name a concrete tell from the artifact, such as a sponsor whose product is the implied solution. The skill aims to reach the primary study behind the input and records a finding when it cannot.
The report ends with a split between what the data supports and what was editorialized, so you can rewrite the claim honestly. A funder's stake is treated as a multiplier on the analysis, not an automatic reason to dismiss the source. Categories with no evidence in the source are left out.
- Layers 1 and 2 are run alone when the input is itself the primary source, with a note that Layer 3 was skipped.
- If the primary study is not publicly accessible, the report records this as publication-availability bias.
What BiasCheck needs and what it does on your machine
BiasCheck requires an agent that supports SKILL.md skills, such as Claude Code. It needs network access to fetch URLs and cited studies. It does not run scripts. An optional Assay-based eval file (EvalState.md) is referenced by the skill but is not included in the provided files.
The skill declares that Edit, Write, and NotebookEdit tools are disallowed. Its only customizations are read from a user folder under your Claude config directory if that folder exists.
- Runs scripts: no
- Uses network: yes, to fetch URLs and cited studies
- Edit, Write, and NotebookEdit are disallowed tools
- Optional user overrides live at <configRoot>/LIFEOS/USER/CUSTOMIZATIONS/SKILLS/BiasCheck/
Tips for better bias checks
Give the skill the most direct link you have. Checking the article alone risks missing whether the underlying data was sound, so a link to the cited study helps. If you only have a pasted abstract, expect a shorter assessment because methodology details may be missing.
Use BiasCheck for questions about evidence and stakes. For broad literature reviews or research synthesis, the skill points you to a separate Research skill. For background checks on companies or people, it does not apply.
- Paste the full abstract or article text when you cannot provide a link.
- Ask who funded a study directly if you want the conflict-of-interest focus.
- Treat the result as an analysis to verify, not a final judgment.
What people use it for
- Checking whether a news story's statistics or headline match the underlying study.
- Reviewing a vendor-funded survey or report before relying on it in a decision.
- Finding out who funded a study and whether the source has a stake in its conclusion.
- Spotting methodological flaws such as unrepresentative samples or leading survey questions.
- Restating an overstated claim in a more accurate form.
Frequently asked questions
Is the BiasCheck skill free?
The source provided for this page lists the licence as MIT. It does not state pricing, so check the licence file in the repository for the terms.
What do I need to use BiasCheck?
You need an agent that supports SKILL.md skills, such as Claude Code, plus network access so the skill can fetch URLs and cited studies. It does not run scripts.
How do I install the BiasCheck skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for all projects or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
Does BiasCheck judge an author's personality?
No. The skill explicitly excludes psychological author analysis. It also excludes entity due diligence on companies or people.
Can BiasCheck do a literature review?
No. Broad literature reviews and research synthesis belong to a separate Research skill.
What if the study the article cites cannot be found?
The skill tries to fetch the primary source. If it cannot be located, the report records that the primary source was not publicly accessible as a publication-availability finding.
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
BiasCheck suits readers, students, and analysts who want to test a news story, report, survey, or vendor study against its underlying evidence and funding. It is not the right tool for judging an author's character, synthesising a research field, or vetting a company or person, and its findings should not replace legal or financial due diligence.
Download the skill