factcheck

factcheck is a skill for Claude Code, Codex from duthaho/skillhub. It costs 213 tokens per session (1,662 once invoked), scanned A, original, MIT.

A citation audit for a document, URL, or draft. It checks whether each factual claim is actually supported by the source linked to it.

In plain words
What is it for?
Use it to extract checkable claims, verify their sources, and audit an entire document or a selected section.
Why use it?
A document can contain working links that do not support its claims. This audit exposes those mismatches and missing evidence.

Skill for Claude CodeCodex

Part of the research plugin — 7 skills shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/duthaho/skillhub/factcheck
Any agent
npx skills add duthaho/skillhub --skill factcheck
Clone the repo
git clone --depth 1 https://github.com/duthaho/skillhub

Made for: Claude Code, Codex.

Or install research, the plugin that ships this one along with the rest of its 7 skills.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for factcheck

README.md
[![agentmods](https://agentmods.dev/badge/skills/duthaho/skillhub/factcheck.svg)](https://agentmods.dev/skills/duthaho/skillhub/factcheck)
Your own site
<a href="https://agentmods.dev/skills/duthaho/skillhub/factcheck"><img src="https://agentmods.dev/badge/skills/duthaho/skillhub/factcheck.svg" alt="Measured on agentmods" height="20"></a>
Per session 213 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,662 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00213 $0.01662
Opus 5 $0.00106 $0.00831
Sonnet 5 $0.00043 $0.00332
Haiku 4.5 $0.00021 $0.00166

Measured 5d ago against content hash 518000eeb594, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

factcheck scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/skills/factcheck/SKILL.md · 144 lines

How it starts

The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.

factcheck — do the sources say that?

/factcheck <path | URL> — audit a document's claims against its sources /factcheck <path> focus <section/topic> — audit only part of a long document

Answer one question: for each claim this document makes, is there a source that actually asserts it? The documented failure mode of cited writing — human and agent alike — is not missing links; it's links that resolve to pages that don't say that. A working URL is the weakest form of evidence there is. This audit replaces "it has citations" with "here is the passage."

Step 0 — Scope the audit

Read the document (WebFetch if a URL; ask for a paste if it's paywalled). Then size the job and say which tier you picked:

  • Small (≤ ~10 checkable claims): verify inline, no sub-agents — the fan-out below would cost more than it saves.
  • Standard: fan out, batching ~5–8 claims per verifier.
  • Focused: the user named a section or topic → extract only there, note in the audit header that the rest was not examined.

If the document has no citations at all, say so up front — the audit's question shifts from "does the cited source say this" to "does support for this exist anywhere", and UNSUPPORTED becomes the default grade to beat.

Step 1 — Extract the checkable claims

Walk the document and number every checkable claim: factual, specific, and in principle verifiable — statistics, quotes, "X said/announced Y", version numbers, benchmark results, historical facts, "studies show".

For each claim record: a short restatement, its locator (section heading or line), the cited source if any, and whether it's date-sensitive (anything with "currently", "latest", a version, or a number that drifts).

Skip pure opinion and analysis — but flag opinion dressed as fact ("it is widely known that...") as a finding in its own right, ungraded.

Step 2 — Adversarial verification (fan out)

Spawn verifier sub-agents (subagent_type: "Explore"), each with a batch of claims and the same standing instruction: try to refute each claim, not confirm it — a verifier hunting for agreement finds it whether it exists or not. Per claim:

Read the full file on GitHub · 144 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 144 lines · 213 tokens per session scan A 518000eeb594

Subscribe to this mod's changes

factcheck is a skill published in the GitHub repository duthaho/skillhub (9 stars, last pushed 12d ago), licensed MIT. It adds 213 tokens to every session and 1,662 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories