researchGrade S · 89/100 Safety 84

Fact Checker

Claim-by-claim verification with verdicts

Splits any text into checkable claims, verifies each against sources, and returns supported / contradicted / unverifiable verdicts with evidence.

claim extractionsource verificationverdict reports

Works with: OpenClaw · Claude Code · Cursor · Any MCP agent · v2.0.1 · MIT · by skillforge-curated

SKILL.md preview

# Fact Checker

## Triggers
User pastes text and asks what is true, or an agent drafts risky content.

## Workflow
1. Split input into atomic checkable claims.
2. Verify each claim against at least 2 sources.
3. Label every claim: supported, contradicted, or unverifiable.
4. Return a verdict table with links.

## Guardrails
- Never launder a guess as verified.
- Show your sources or say none found.

Install in one command

1. Create the skill folder: mkdir -p skills/fact-checker

2. Install: curl -fsSL "https://skillforge-aniruddha-adaks-projects.vercel.app/api/skills/fact-checker/skill.md" -o skills/fact-checker/SKILL.md

3. Mention it in SOUL.md or BOOTSTRAP.md so OpenClaw loads it on boot.

Install command

curl -fsSL "https://skillforge-aniruddha-adaks-projects.vercel.app/api/skills/fact-checker/skill.md" -o skills/fact-checker/SKILL.md

Why this score

Documentation16/20

3/3 required SKILL.md sections, 402 chars

Capability breadth12/15

3 declared capabilities

Agent compatibility15/15

Works with openclaw, claude-code, cursor, generic-mcp

License openness10/10

MIT — reuse freely

Community signal20/20

12900 registry installs

Safety posture16/20

Risk surface: network-egress

Audit seal

SHA-384(prev ‖ canonicalJson)

5ebe1112841a94fd81017eb9022f1dca14ee481e2671f65281a726ec19f4bdc8e195efa6754611a7148f69b0b79f23fd

chain head e800631e36483c866a9a2ba7ae87b272…

Replay: POST {ids} to /api/seals/verify

Pairs well with

Deep Researcher92/100 Paper Scout89/100 HN Pulse88/100