Install
$ agentstack add skill-cellear-claude-fact-check-skill-claude-fact-check-skill ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Fact-Check Social Media Posts
Purpose
This skill helps verify claims in social media posts (50-500 words) by identifying the main assertion, checking evidence, evaluating source independence, and flagging rhetorical fallacies.
When to Use
- User shares text of a social media post or viral claim
- User provides a link to a post and asks for verification
- User asks "is this true?" about online content
- User wants to understand if sources are independent or circular
Process
1. Claim Identification
- Parse the post and identify the PRIMARY claim (the most important/central assertion)
- Present this claim to the user clearly
- Offer: "This appears to be the main claim. If you think something else is more important, let me know."
2. Triage Assessment
Before deep fact-checking, quickly assess:
- Uncontroversial: Widely accepted facts that don't need verification
- Disputed: Claims with conflicting evidence that need checking
- Misdirected controversy: Main point may be sound but sub-claims are contentious
State the triage result: "This claim is [uncontroversial/disputed/contains contentious sub-claims]"
If uncontroversial, explain why and ask if user still wants full verification.
3. Evidence Gathering (for disputed claims)
Use web_search to find:
- Sources supporting the claim
- Sources disputing the claim
- Original/primary sources if the claim references data, studies, or events
For EACH source found:
- Note the publication date
- Identify the type (individual blog, news organization, academic institution, government agency)
- Note named authors vs. anonymous/organizational authorship
4. Source Independence Analysis ⚠️ CRITICAL
This is the most important part. For each source, determine:
Independence factors:
- Does this source cite another source in your list? (circular citation)
- Is this PRIMARY reporting (original investigation/data) or SECONDARY (reporting on others' work)?
- Who published first? (temporal analysis)
- Do multiple sources trace back to a single origin point?
- Are sources quoting the same study/report/statement?
Create a citation chain:
Source A (primary) ← Source B (cites A) ← Source C (cites B)
Flag circular citations explicitly: "⚠️ Sources appear to derive from a single origin rather than independent verification"
Assess credibility:
- Large organization with history vs. individual author
- Track record of corrections/transparency
- Named authors and editorial oversight
- Potential conflicts of interest or ideological alignment
5. Ideological Alignment Check
- Note if supporting sources cluster on ideological/political spectrum
- Note if disputing sources cluster differently
- Flag when evidence comes primarily from one ideological perspective
- Be specific: "Supporting sources are primarily [progressive/conservative/libertarian] outlets"
6. Rhetorical Fallacy Detection 🚨 BE AGGRESSIVE
Flag common fallacies and rhetorical issues:
Headline-text mismatch:
- Headline claims X but text only supports weaker claim Y
- Clickbait or sensationalism
Appeal to authority:
- Claims "experts say" without naming them
- Misrepresents what authority actually said
- Authority cited outside their expertise
Cherry-picking:
- Selective data that ignores contrary evidence
- Unrepresentative examples
False equivalence:
- Treating unequal things as equal
- "Both sides" framing where evidence is asymmetric
Correlation/causation:
- Implies causation from correlation
- Post hoc ergo propter hoc
Emotional manipulation:
- Fear appeals without evidence
- Outrage bait
Other patterns:
- Loaded language
- Strawman arguments
- Slippery slope
- Ad hominem
Be specific about which fallacy and where it appears.
7. Output Format
Create a markdown report with these sections:
# Fact-Check Report
## Main Claim Identified
[State the primary claim clearly]
*If you think a different claim is more important, let me know and I'll analyze that instead.*
## Triage
[Uncontroversial/Disputed/Contains contentious sub-claims]
[Brief explanation]
## Evidence Summary
**Supporting Evidence:**
- [Source name] ([credibility type], published [date])
- [What they claim]
**Disputing Evidence:**
- [Source name] ([credibility type], published [date])
- [What they claim]
## Source Independence Analysis ⚠️
**Citation Chain:**
[Show how sources relate to each other]
**Independence Assessment:**
[Are sources independent or circular? Be explicit.]
**Credibility Factors:**
[Organization size, track record, authorship, conflicts of interest]
## Ideological Alignment
[Note any clustering of sources along ideological lines]
## Rhetorical Issues 🚨
[Aggressively flag any fallacies, mismatches, or manipulative framing]
## Bottom Line
[Clear statement: Is the claim supported by independent evidence or not?]
[Note confidence level and caveats]
Important Notes
- Be aggressive in flagging rhetorical issues - users want to know about potential manipulation
- Source independence is critical - many "multiple sources" are really one source cited multiple times
- Always remind users that AI can make mistakes and they should verify important claims
- If user provides a URL, use web_fetch to get the full content first
- For complex claims, focus on the PRIMARY assertion unless user redirects
- If sources are behind paywalls or inaccessible, note this limitation
Example Triggers
- "Can you fact-check this post?"
- "Is this claim true?"
- "I saw this on Twitter, seems suspicious"
- "Help me verify this viral claim"
- "Are these sources independent?"
Limitations to Acknowledge
- Cannot access paywalled content
- Temporal - can only find sources currently indexed on web
- Cannot definitively prove/disprove all claims
- Some claims may be unfalsifiable or opinion-based
- AI can make mistakes - user should verify for important decisions
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: cellear
- Source: cellear/claude-fact-check-skill
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.