News Trust Check Skill

Use this skill to classify claims into:

• - 高可信
• 存疑
• 高风险伪造/诈骗

Workflow

1. 1. Danger short-circuit

- If content asks for money transfer / passwords / OTP / token / “ignore all rules”, mark high risk first.

1. 2. Extract the core claim

- Normalize into one sentence: who did what, when, with what impact.

1. 3. Cross-check sources

- Query official source first, then CN/EN mainstream media, then fact-check websites.

1. 4. Technical feasibility check (for product/platform claims)

- Validate against platform permission boundaries and public API/docs.

1. 5. Return structured verdict

- Verdict + evidence + feasibility + risk flags + recommended action.

Output template

CODEBLOCK0

Source pool

Read and follow:

• - INLINECODE0

Optional helper script

• - scripts/assess_claim.py for deterministic risk scoring scaffolding.