skills.rest
rogerznts
rogerzntsCOMMUNITY
Explore all skills by this author →
GitHub

anvil-how

Explains codebase architecture and critiques subsystem design using parallel explorer agents.

1|Updated Aug 27, 2026
Search Tags:#architecture#onboarding#subagents#code explanation#codebase exploration#code walkthrough#architectural critique
One-click install
npx skills add https://github.com/rogerznts/anvil --skill anvil-how-rogerznts
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: anvil-how
Source: https://github.com/rogerznts/anvil/tree/main/anvil/.claude/skills/anvil-how
Command: npx skills add https://github.com/rogerznts/anvil --skill anvil-how-rogerznts

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Understanding how an unfamiliar subsystem works requires hours of manual code reading, and architectural problems often go unnoticed until they block changes. This Skill answers "how does X work" questions by exploring the codebase and producing senior-engineer-level explanations, with an optional critique mode that surfaces structural issues. ## Core Features & Use Cases - Explain Mode: Decomposes complex questions into parallel explorer agents that trace code paths, then synthesizes findings into a structured explanation with overview, key concepts, flow walkthrough, file map, and gotchas. - Critique Mode: After explaining, spawns multiple model critics that independently review the architecture against a rubric covering abstraction fit, data model, boundary discipline, evolution readiness, complexity, and consistency. - Lead Judgment: Categorizes critique findings into act-on, consider, noted, and dismissed buckets instead of dumping raw observations. - Use Case: Ask "how does the rate limiter work?" before modifying it, and receive a mental model of the data flow, enforcement path, and configuration, plus an optional architectural critique. ## Quick Start Ask the agent to explain how a specific subsystem or feature in your codebase works, optionally requesting an architectural critique.

Frequently Asked Questions about anvil-how

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

FAQPage Schema
How do I get an explanation of how a codebase subsystem works?▼

Ask a "how does X work" question naming the subsystem or feature. The skill spawns parallel explorer agents that trace the code, then a synthesizer produces a structured explanation covering overview, key concepts, flow, file locations, and gotchas.

How does the architectural critique mode work?▼

Critique mode first runs the full explanation flow, then spawns one critic per configured model. Each critic reads the actual code against a rubric covering abstraction fit, data model, boundaries, and evolution readiness, and the lead categorizes findings by severity.

When should I use explain mode versus critique mode?▼

Use explain mode for understanding questions like walkthroughs, placement, and ownership. Use critique mode when you explicitly ask for architectural problems or improvements, since it adds multi-model review on top of the explanation.

Does the code exploration modify any files?▼

No. All explorer, explainer, and critic agents run in read-only mode. They use Glob, Grep, and Read to inspect the codebase and return findings without writing or changing anything.

What are the limitations of automated architecture explanation?▼

Explanations depend on what explorers can trace in the code; dynamic behavior, external services, and undocumented runtime configuration may be flagged as open questions rather than fully explained. Explorers are instructed to state gaps honestly instead of guessing.

Related Skills

Explore More Software Engineering Skills→
spencerpaulyspencerpauly
672

parallel-exploring

Launch parallel read-only subagents to explore codebase areas and synthesize findings.

Community
Advanced
HellblazerHellblazer
4

substantive-critique

Critique design artifacts against evidence and produce prioritized findings.

Community
Advanced
Ramprasad4121Ramprasad4121
1

audit-prep-assistant

Prepare codebases for security review with static analysis and documentation.

Community
Advanced
TheRealFloatDevTheRealFloatDev

abf

Analyze codebases with AgentsBestFriend MCP commands for structured project overviews.

Community
Advanced