advisor-workflow

Coordinates implementation work with advisor subagent consults for design-bearing code changes.

2.0k|99|Updated Feb 15, 2019
One-click install
npx skills add https://github.com/reactive/data-client --skill advisor-workflow
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: advisor-workflow
Source: https://github.com/reactive/data-client/tree/main/.cursor/skills/advisor-workflow
Command: npx skills add https://github.com/reactive/data-client --skill advisor-workflow

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Deciding when a cheap, fast implementing model should consult expensive advisor subagents is hard: consult too often and you waste cost and latency, consult too rarely and you ship flawed designs that are expensive to reverse. This Skill provides a cost-and-speed-optimized workflow that routes design-bearing or high-risk implementation decisions to the right advisor subagent at the right time.

Core Features & Use Cases

  • Triage and risk routing: Classifies tasks as mechanical, design-bearing, or high-uncertainty, with overrides for security, concurrency, compatibility, and previously failed attempts.
  • Structured advisor consults: Routes decisions to design-advisor, principal-advisor, and quality-reviewer subagents using a standardized context packet format (DECISION, ALTERNATIVES, CONSTRAINTS, EVIDENCE, and more), with parallel batching and blocking-vs-background modes.
  • Runtime validation and escalation: Requires build/test/browser evidence, defines an escalation valve after repeated failed hypotheses, and mandates an end-of-implementation diff review with disposition tracking.
  • Use Case: When adding a new public API schema to a repository, the implementer extracts the contract decisions, consults design-advisor with a context packet, implements against the decided interface, validates with real test runs, and sends the final diff to quality-reviewer before completion.

Quick Start

Invoke the advisor-workflow skill before implementing a change that touches public interfaces, schemas, concurrency, or security so decisions are reviewed by advisor subagents.

Frequently Asked Questions about advisor-workflow

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

FAQPage Schema
How do I decide when to consult an advisor subagent during implementation?

Consult when a decision carries high rework cost over the project's lifetime, such as new public signatures, schemas, or cross-module contracts. Skip consults for mechanical edits like renames, config changes, or well-specified small fixes.

What is the difference between design-advisor and principal-advisor subagents?

design-advisor is a mid-price model for interface, API, schema, and data-flow decisions, used roughly twice per task. principal-advisor is an expensive, slow model reserved for the single highest-blast-radius decision such as architecture, concurrency, or security.

When should a principal-advisor consult run in the background?

Run it in the background only when substantial work is genuinely independent of its answer. If the decision gates the rest of the plan, launch it blocking, since implementing against a guess risks rework larger than the wait.

What should I do after two failed attempts at the same problem?

Stop and never repeat an approach without new evidence. Package the failure evidence, including error output and why each attempt failed, and consult design-advisor, or principal-advisor if the failure is correctness-deep.

Is an end-of-implementation code review always required?

Review by quality-reviewer is mandatory for design-bearing or high-risk work, and for any diff touching lifecycle, public APIs, or three or more files. An unresolved critical finding blocks completion and escalates to principal-advisor.