Skip to content
Back to skills

Verify

ASecurity

Fact-check claims encountered during reading — dates, names, events, citations. Use when encountering historical facts or disputed claims.

  • 41 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 3, 2026
researchgodatabase

Security analysis

A100/100

Scanned September 3, 2026

npx -y skills add yha9806/academic-writing-toolkit --skill verify --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Verify?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Verify
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/yha9806-verify/badge)](https://www.skillsdirectory.com/skills/yha9806-verify)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: verify
description: Fact-check claims encountered during reading — dates, names, events, citations. Use when encountering historical facts or disputed claims.
allowed-tools: WebSearch, WebFetch, Read, Edit
---

# /verify — Fact Verification Skill

## Purpose

Fact-check claims encountered during reading sessions. This covers dates, names, historical events, publication details, and attribution of ideas. Use when the user questions a claim or when a factual assertion seems uncertain.

## Trigger Words

This skill activates on: `verify`, `check this`, `fact-check`, `is this correct`, `/verify`.

## Workflow

1. **Extract the claim** from the user's input or from conversation context. Identify the specific factual assertion to verify (date, name, event, publication, attribution).
2. **Search authoritative sources** using WebSearch. Prioritise:
   - Wikipedia (for dates, biographical facts, historical events)
   - Academic databases (Google Scholar, JSTOR, PhilPapers)
   - Publisher websites (for publication dates, editions, ISBNs)
   - Encyclopedias (Stanford Encyclopedia of Philosophy, Britannica)
3. **Compare** the original claim against verified information.
4. **Output the verdict** using the format below.
5. **Optionally annotate the notes file.** If the user confirms, use Edit to add a `(verified)` or `(corrected: {correct info})` tag next to the relevant entry in the notes file.

## Output Format

```
## Verification

**Claim**: {the original claim as stated in the source}
**Verdict**: Confirmed | Needs correction | Incorrect | Unverifiable

| Item | Original | Verified |
|------|----------|----------|
| {specific fact} | {as claimed} | {as verified} |

**Source**: {single-line citation in the project's declared style — see `literature/reading_notes/_template_NOTES.md` for per-style examples; the active style is `Citation style:` in `CLAUDE.md`. Append the URL after the citation if the source is web-only}
**Confidence**: High | Medium | Low
**Notes**: {any additional context, e.g., "Date varies by edition" or "Multiple conflicting sources"}
```

## Multiple Claims

When verifying several claims at once, produce one verification block per claim, numbered sequentially:

```
### 1. {Brief claim description}
{verification block}

### 2. {Brief claim description}
{verification block}
```

## Annotation

When the user says "annotate" or "mark in notes" after verification:
- Use Edit to find the relevant passage in the notes file.
- Append the tag inline: `(verified 2026-03-30)` or `(corrected: {correct value}, verified 2026-03-30)`.
- Do not alter the original text of the note -- add the tag after it.

## Constraints

1. **Never auto-correct notes** without user confirmation. Present the verdict first, then ask if the user wants to annotate.
2. **Cite sources** with URLs whenever possible. Do not present unverifiable assertions as fact.
3. **Acknowledge uncertainty.** If sources conflict, report all versions and mark confidence as Low.
4. **No emoji** in output.
5. **Scope**: This skill verifies factual claims only. For argument analysis or interpretation, defer to the user's own reading.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.