decision-guard

Anchor decision questions to the canonical GraphReFly Rust-port decision log.

1|Updated Mar 29, 2026
One-click install
npx skills add https://github.com/graphrefly/graphrefly-ts --skill decision-guard
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: decision-guard
Source: https://github.com/graphrefly/graphrefly-ts/tree/main/.claude/skills/decision-guard
Command: npx skills add https://github.com/graphrefly/graphrefly-ts --skill decision-guard

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

decision-guard prevents silent drift by anchoring answers to the user’s locked GraphReFly Rust-port decisions, values, invariants, and the canonical D-numbered decision log so new proposals don’t conflict with prior scope or architecture choices.

Core Features & Use Cases

  • Canonical decision recall: Loads immovable user values/principles and the locked decision-process patterns for consistency.
  • Spec-first consistency enforcement: Directs you to treat the canonical Rust-port spec as authority and to halt when the spec is ambiguous or out of date.
  • Safe decision triage guidance: Helps classify whether a proposed change is consistent, requires a new D-number (scope formalization), or should be rejected silently due to prior deferrals/divergences.

Quick Start

Use the decision-guard skill before answering any question that asks whether something is consistent with prior decisions, scope, or a proposed fix.

Frequently Asked Questions about decision-guard

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

FAQPage Schema
How do I prevent scope drift when porting to Rust across multiple sessions?

Spec validation requires loading the canonical Rust-port spec as the primary authority source. You must validate all proposed changes against this spec first, halting immediately if the spec is ambiguous or out of date.

How do I enforce decision consistency during triage and regression discussions?

Decision consistency is enforced by classifying proposed changes against locked values and the D-numbered decision log. Changes are marked consistent, requiring a new D-number for scope formalization, or rejected due to prior deferrals.

What are the HALT conditions for autonomous scope expansion in a multi-agent workflow?

HALT conditions trigger for autonomous scope expansion, stale premises, async-at-boundary violations, and cross-track-ledger widening without the required ledger update. These prevent silent drift during multi-agent workflows.

When do I need to create a new D-number for a proposed change?

A new D-number is required for scope formalization when a proposed change is classified as needing formalization during triage. This occurs when the change expands beyond previously locked decision boundaries.

Can I use decision-guard for option selection like A/B/C or alpha/beta/gamma comparisons?

Yes, decision-guard applies to option selection during porting, triage, and regression discussions. It validates A/B/C, X/Y/Z, and alpha/beta/gamma options against the canonical decision log to prevent silent drift.