nbauthorize-2-stage-decisions

Gate LLM claim extraction with two-stage NBAccess send and persist decisions.

3|Updated Mar 17, 2026
One-click install
npx skills add https://github.com/transreal/claudecode --skill nbauthorize-2-stage-decisions
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nbauthorize-2-stage-decisions
Source: https://github.com/transreal/claudecode/tree/main/Claude%20Directives/skills/nbauthorize-2-stage-decisions
Command: npx skills add https://github.com/transreal/claudecode --skill nbauthorize-2-stage-decisions

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It prevents unsafe or non-compliant extraction by enforcing NBAccess authorization both before sending sensitive content to an LLM and before persisting extracted claims, while minimizing extra authorization cost.

Core Features & Use Cases

  • Two-phase authorization for extraction: Performs a send-stage decision and a persist-stage decision using SourceVaultExtract so that content is authorized for LLM transmission and for claim storage.
  • Extended decision handling: Supports four decision outcomes—Permit, Screen, RequireApproval, and Deny—with fail-safe behavior that treats unknown decisions as denied.
  • Fail-safe response shaping: Returns Status values like DeniedByNBAccess and RequiresApproval, and includes AccessDecisions as <|"Send" -> _, "Persist" -> _|> (or an empty association when explicitly opted out).
  • Correct claim spec construction from source metadata: Builds NBClaimSpec via iSpecFromClaim, deriving AccessLabel from snapshot/source meta so NBAccess can apply consistent labeling rules.

Quick Start

Request the system to run SourceVaultExtract with the default authorization enabled for sensitive sources so it returns extracted claims only when both send and persist NBAccess checks succeed, otherwise returning DeniedByNBAccess or RequiresApproval with AccessDecisions details.

Frequently Asked Questions about nbauthorize-2-stage-decisions

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I enforce authorization checks before sending data to an LLM and before saving extracted claims?

You can enforce two-stage LLM authorization by applying send and persist decision gates, ensuring content is checked for compliance before transmission and again before claim storage.

What happens to LLM extracted claims if an authorization decision is unknown or unsupported?

Unknown or unsupported authorization decisions for LLM extracted claims are treated as denied by default, returning a fail-safe status to prevent non-compliant data persistence.

Can I batch authorize LLM extraction requests for multiple sensitive sources?

Yes, you can batch authorize LLM extraction requests by using a representative claim to evaluate the send and persist decisions for the entire group of sensitive sources.

Does the two-stage authorization approach support outcomes beyond simply permit or deny?

Two-stage LLM authorization supports four decision outcomes—Permit, Screen, RequireApproval, and Deny—allowing nuanced governance like returning a RequiresApproval status before claim persistence.

How do I disable LLM governance checks for a SourceVault extraction flow?

You can disable LLM governance checks by using the opt-out switch, which yields empty AccessDecisions and bypasses the send and persist authorization gates during the extraction flow.