skills.rest
avdi
avdiCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

auditor-family

Bootstrap and maintain a family of standards skills and auditor agents for project code review.

Updated May 9, 2020
Search Tags:#code review#subagents#software quality#ai workflow#review framework#auditor agents#standards skills
One-click install
npx skills add https://github.com/avdi/purse --skill auditor-family-avdi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: auditor-family
Source: https://github.com/avdi/purse/tree/main/home/dot_agents/skills/auditor-family
Command: npx skills add https://github.com/avdi/purse --skill auditor-family-avdi

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Agent-based code review decays quickly: rules get duplicated between guidelines and reviewers, reviewers get dispatched on irrelevant changes, and design-level smells slip through because no checklist owns them. This Skill provides a complete framework for setting up and maintaining a family of paired standards skills and auditor agents that review code consistently and evolve with the project. ## Core Features & Use Cases - Framework bootstrapping: Discover review areas along three axes (modules, cross-cutting concerns, technology), then write paired standards skills and area auditors using provided templates. - Structured review cadence: Run a one-round spec audit before implementation and a capped two-round diff audit, with an auditor-general dispatching tiered subagents and a generalist closing pass catching design smells no area owns. - Continuous evolution: Use curator and toolmonger meta-agents to harvest durable lessons from reviews into standards and turn ad-hoc commands into reusable tools. - Use Case: When setting up agent code review for a new repository, use this Skill to write the project brief, discover auditable areas, generate the standards skills and auditor agents, and wire the review workflow into AGENTS.md. ## Quick Start Use the auditor-family skill to bootstrap an agent code review framework for this repository, starting with the project brief and area discovery.

Frequently Asked Questions about auditor-family

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

FAQPage Schema
How do I set up agent-based code review for a repository?▼

Write a project brief first, then discover review areas along three axes: modules, cross-cutting functional concerns, and technology stacks. For each area, write a standards skill with concrete rules before creating its paired auditor agent, then add an auditor-general dispatcher.

What is the difference between a standards skill and an auditor agent?▼

A standards skill holds the concrete coding rules with bad/good examples and is loaded before doing work. An auditor agent enforces that skill during review, carrying only values, direction, and severity calibration while citing the skill rather than restating rules.

How many review rounds should an agent code audit run?▼

The framework caps diff audits at two rounds. Round 2 is scoped to a commit range bounding only the fixes, and it is final, so auditors must not hold findings back expecting a third round.

When should I split an auditor into multiple auditors?▼

Split an auditor when its charter contains two sets of unrelated invariants, and split the paired standards skill with it. Do not split merely because files differ; split when the rules themselves diverge.

Why does my review framework keep missing design problems?▼

Area auditors only see their own slice, so agglomeration and pattern drift fall between them. The generalist closing auditor reads the full diff plus the consolidated verdict with fresh eyes specifically to catch design and architecture smells no written rule covers.

Related Skills

Explore More Software Engineering Skills→
LeoCode03LeoCode03

setup

Configure the Praxis skills pack by detecting project stack and writing CLAUDE.md and ARCHITECTURE.md.

Community
Advanced
tta-labtta-lab
23

ttal-comment

Coordinate review comments and approvals across TTal pipeline stages.

Official
Intermediate
schattenspiegelschattenspiegel

github-copilot-custom-agents

Create and verify GitHub Copilot custom agents with least-privilege tool restrictions and handoffs.

Community
Intermediate
yang-lyang-l

design-doc-full-pass

Orchestrates parallel read-only agents to edit large design docs section by section.

Community
Advanced