skills.rest
austintheriot
austintheriotCOMMUNITY
Explore all skills by this author →
GitHub

agent-modes

Standardize output formats for review, plan, and consult agent workflows.

1|Updated Jul 31, 2023
Search Tags:#code review#multi-mode#agent modes#spec critique#expert consulting#output contracts#severity schema
One-click install
npx skills add https://github.com/austintheriot/dotfiles --skill agent-modes-austintheriot
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-modes
Source: https://github.com/austintheriot/dotfiles/tree/main/.claude/skills/agent-modes
Command: npx skills add https://github.com/austintheriot/dotfiles --skill agent-modes-austintheriot

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates ambiguity in multi-mode expert agents by defining the exact output contract expected for review, plan, and consult workflows.

Core Features & Use Cases

  • Review mode contracts (diff/survey): Produces findings in the required panel-contract severity, confidence, and anchoring format, with explicit guidance on what to ignore.
  • Plan mode contracts: Critiques a specification without inventing code, identifying design gaps, missing requirements, assumptions, and risks using the severity vocabulary.
  • Consult mode contracts: Provides direct expert answers in the shape the question demands, using references from the loaded rules lens while avoiding review/spec critique scaffolding.
  • Mode detection rules: Establishes how to infer the intended mode when it is not explicitly named, reducing dispatch mistakes.

Quick Start

Instruct your agent to use the correct mode contract for the dispatch prompt it receives (review with diff or survey, plan for spec critique, or consult for question answering).

Frequently Asked Questions about agent-modes

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

FAQPage Schema
How do I standardize AI agent output formats for code review and spec critique?▼

Standardize AI agent output formats by defining explicit output contracts for review, plan, and consult workflows. This ensures consistent severity vocabulary and anchoring format across multi-mode expert agents.

What is the difference between diff and survey scopes in code review agent modes?▼

Diff and survey scopes in code review agent modes distinguish between reviewing specific code changes versus evaluating broader existing code regions. The output contract enforces explicit handling for each scope to maintain review focus.

How do I configure consult mode to prevent unwanted review scaffolding in agent responses?▼

Configure consult mode by applying output contracts that enforce clear separation of concerns, disallowing review scaffolding and linter duplication. This ensures the agent provides direct expert answers strictly in the demanded shape.

How do agents infer the correct workflow mode when it is not explicitly named?▼

Agents infer the correct workflow mode by applying built-in mode detection rules to the dispatch prompt. These rules analyze the prompt context to reduce dispatch mistakes between review, plan, and consult workflows.

What should an agent ignore when performing a plan mode specification critique?▼

In plan mode specification critique, an agent should ignore inventing code and instead focus on identifying design gaps, missing requirements, and risks. It uses the defined severity vocabulary to evaluate the specification.

Can I use a single agent for both direct Q&A and code specification evaluation?▼

Use a single multi-mode agent for both direct Q&A and code specification evaluation by switching between consult and plan modes. Each mode applies distinct formatting rules to separate direct answers from structured spec critique findings.

Related Skills

Explore More Software Engineering Skills→
luongnv89luongnv89
40.9k

brand-voice-consistency

Check content against brand voice and tone guidelines.

Community
Basic
atomicinnovationatomicinnovation
30

plan-review-output-format

Define a standardized JSON output format for plan review agents.

Official
Basic
padjeipadjei

design-systems-slds-validate

Validate Lightning Web Components for SLDS compliance and generate a detailed scorecard.

Community
Advanced
charly-vibescharly-vibes

extract-prompt

Extract reusable prompt templates from AI conversations by analyzing successful patterns.

Community
Intermediate