Skill Details
Runtime
(claude-5-opus)
Example user message
[userMessage]: Our support RAG answers: “Enterprise customers can request a 30-day data retention period.” The retrieved evidence contains: 2026 Enterprise Data Policy: retention is configurable from 7–30 days. 2025 Security FAQ: enterprise retention is fixed at 7 days. 2026 policy metadata says it supersedes the 2025 FAQ. Check whether the answer is grounded and provide the safest final response.
Example agent response
Evidence Assessment
The claim is supported by the supplied 2026 Enterprise Data Policy and its stated supersession metadata.
Authoritative evidence
The 2026 Enterprise Data Policy states that retention is configurable from 7–30 days.
Its metadata states that it supersedes the 2025 Security FAQ.
Conflicting evidence
The 2025 FAQ states a fixed 7-day period, but the supplied metadata identifies that document as superseded.
Claim Check
Answer claim
Evidence
Support
Enterprise customers can request a 30-day retention period
2026 Enterprise Data Policy
Directly supported
Retention is configurable
2026 Enterprise Data Policy
Directly supported
7-day retention is mandatory
2025 Security FAQ
Contradicted by the superseding policy
Grounded Answer
Enterprise customers can configure data retention for 7–30 days, so a 30-day retention period is supported by the current 2026 Enterprise Data Policy. The older 7-day statement comes from a superseded FAQ and should not be treated as the current policy.
Decision: SUPPORTED — based on the supplied authoritative 2026 source and supersession metadata.
Get this skill + 9 more with Select$14 first month, then $19/mo · Cancel anytime$14 first month · then $19/mo
View Select
Instant access
Commercial use
Money‑backBy purchasing this skill, you agree to our terms of service
4
Views
CLAUDE-5-OPUS
Build reliable RAG systems that retrieve and verify relevant evidence, preserve source context, ground answers in retrieved content, and detect unsupported claims. Design query, retrieval, reranking, context assembly, citation or attribution, and answer-generation workflows with measurable evaluation, failure analysis, regression checks, and traceable change impact. Optimize quality, latency, and cost while validating retrieval and answer faithfulness across real-world edge cases.
...more
Added 4 days ago
