Scientific Claim Audit

VerifiedSafe

Audit claims, causal chains, and conclusions in manuscripts or design docs for scientific rigor.

Sby Skills Guide Bot
DocumentationAdvanced
7/25/2026
Claude Code
#claim-audit#scientific-writing#causal-reasoning#overclaiming#manuscript-review

Recommended for

Automated summary

Audits the claims, causal chains, and conclusions of a manuscript or design documents for scientific rigor.

Strengths

  • Extracts each claim verbatim with its location and rates the available support as strong, mixed, weak, or absent.
  • Flags specific overclaiming patterns: "proves" where "suggests" is warranted, universal claims from few examples, unfalsifiable assertions, and mechanism-as-proof reasoning.
  • Tests sufficiency and uniqueness claims against the project's design principles (governance ordering, matching quality dimensions to the number of mechanisms).
  • Delegates the claim-extraction methodology to a dedicated skill and cross-checks robustness claims against the attack-and-defense log.

Limitations

  • Tightly coupled to the decentralized-curation manuscript: its targets (paper.qmd, projects/*/design.md, context/attack-and-defense-log.md) and guiding principles do not transfer unchanged to other projects.
  • Does not verify facts or literature; it judges whether the evidence presented supports the claim, not whether that evidence is correct.
  • Returns a list of claims to revise but does not rewrite the prose or supply new evidence.
When to use it

Use it before finalizing any section that makes empirical or causal assertions, or whenever you suspect overclaiming in a manuscript or design document.

When not to use it

Avoid it for pure copy-editing, citation checking, or auditing text that carries no argumentative claims.

Security analysis

Safe
Quality score82/100

The skill is a pure read-and-reason analysis workflow. Its declared tools are Read, Grep, Glob, and Agent — no shell, network, write, or delete capability — so it cannot exfiltrate data or alter the repository. The only delegation is to another named reasoning skill (claude-scientific-writer:scientific-critical-thinking) as a methodology reference, which is transparent and consistent with the declared Agent tool. No obfuscated payloads, credential handling, or safety-disabling instructions are present.

No concerns found

How this analysis is made

Examples

Audit causal claims in the paper
Audit the causal claims in paper.qmd: for each one, quote it with its location, state what would need to hold for it to be true, rate the support as strong/mixed/weak/absent, and suggest a revision.
Check design docs for sufficiency claims
Check projects/*/design.md and projects/*/blueprint.md for uniqueness and sufficiency claims that aren't justified by the two mechanisms we actually have.
Red-team robustness claims
Red-team our robustness claims against context/attack-and-defense-log.md before I merge this section.

name: paper-audit-claims description: "Audit claims, causal chains, and conclusions in the manuscript or project design docs for scientific rigor. Use when the user asks to check claims, verify arguments, test causal chains, check for overclaiming, or before finalizing any section that makes empirical or causal assertions." allowed-tools: [Read, Grep, Glob, Agent] context: inherit

paper-audit-claims: claim audit for the decentralized-curation manuscript

When to use

Trigger when the user asks to audit claims, check for overclaims, verify causal chains, test assertions, or evaluate whether an argument is supported by the evidence and mechanisms actually available. Also trigger as a pre-merge check on any prose change that adds or strengthens a causal claim.

Delegate methodology to

claude-scientific-writer:scientific-critical-thinking for the claim-extraction and evidence-sufficiency workflow.

Project-specific targets

  • Causal claims about incentive alignment and mechanism-as-proof reasoning in paper.qmd.
  • Uniqueness / sufficiency claims in projects/*/design.md and projects/*/blueprint.md.
  • Red-team robustness claims: cross-check against context/attack-and-defense-log.md, which tracks integrated attacks and defenses.

Project Design Principles to uphold

  • NoNeedForGovernance > Good Governance > Bad Governance > No Governance. Flag designs that justify added governance without ruling out the simpler "no governance needed" case.
  • Do not decompose quality further than your mechanisms can distinguish. Two mechanisms (challenge-based accuracy, coherence-based relevance) mean two dimensions. A third dimension without a third mechanism is noise. (Corollary of Wang and Strong's "decompose quality before you mechanize it.")

Writing-quality signals to flag

  • "Proves" when "suggests" is warranted.
  • Universal claims from limited examples.
  • Unfalsifiable assertions.
  • Mechanism-as-proof claims: describing design as if design proves correct behavior.
  • Appeals to absence of evidence.

Output shape

For each claim: (1) the claim verbatim with location; (2) what would need to hold for it to be true; (3) whether support is strong / mixed / weak / absent; (4) suggested revision or evidence needed.

From the file to a first result

The download contains exactly the content displayed on this page. This copy may differ from the latest version in the source repository.

View the source and supporting files

This download contains SKILL.md only. Any scripts, references, templates and dependencies are not included. Get the required files from the source, preserving their directory structure, before continuing.

1. Download the file

Download SKILL.md

From the root of a test project, save the file as .skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md, or use the command below. This folder is for inspecting the skill before installation.

Command for macOS/Linux or WSL: requires a POSIX shell and curl. On Windows without WSL, use the download button and create the folder manually.

mkdir -p '.skill-downloads/scientific-claim-audit-mrzxupmn' && curl --fail --show-error --location 'https://skills-guides.com/api/skills/scientific-claim-audit-mrzxupmn/download' --output '.skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md.part' && mv '.skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md.part' '.skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md'

2. Check prerequisites

Read SKILL.md before using it. If it references scripts or external services, inspect them and prepare the required access in your test environment.

No recognized prerequisites section in this file. This does not mean it is self-contained: libraries, tools and access requirements may be mentioned elsewhere in the content or source repository.

The platforms on this page are catalog indications. The paths below follow tool documentation; they do not prove this skill has been run in those tools. No minimum tool version is confirmed for this listing.

3. Load the skill

After inspection, copy the complete folder into your project using one of the commands below, then open a new tool session. The command stops if a folder with the same name already exists: compare the files before replacing it.

Claude Code
test -f '.skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md' && test -s '.skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md' && mkdir -p '.claude/skills' && test ! -e '.claude/skills/paper-audit-claims' && cp -R '.skill-downloads/scientific-claim-audit-mrzxupmn' '.claude/skills/paper-audit-claims'

OpenSkills

OpenSkills must already be installed. The local install/read procedure is checked with OpenSkills 1.5.0, which requires Node.js ≥ 20.6.0. Catalog slugs are not resolved automatically.

Official OpenSkills installation
openskills install './.skill-downloads/scientific-claim-audit-mrzxupmn'

openskills read 'scientific-claim-audit-mrzxupmn'

Reading should display the installed file content and its base directory. It does not run the skill or install its dependencies.

For an agent that reads AGENTS.md, then sync the index. This command changes AGENTS.md: review the diff before using it.

openskills sync

Manual method or another tool: attach SKILL.md to a conversation, or open it in an agent that can read files. Include any required supporting resources. No universal folder is recognized by every tool.

4. Try a first task

Use a short, synthetic input: an API endpoint, a small document or an interface brief. In a new conversation, attach the file or give the agent its path, then adapt this request:

Read .skill-downloads/scientific-claim-audit-mrzxupmn/SKILL.md and apply its instructions to this example: [describe a test input]. Identify any missing files, dependencies or access. Produce the result requested by the skill without inventing data or running unreviewed scripts.

Expected result: output matching the skill instructions, or a precise list of missing prerequisites. A read confirmation alone does not validate a first use. If an external resource is unavailable, that path remains untested.

Related skills