Archify Tuning Playbook

An internal, evidence-driven loop for improving Archify diagram accuracy, readability and authoring speed.

Dev & CodingAdvanced★ 91,820⑂ 6,257AI score 6/10Last updated: Oct 30, 2026

What it does

Locks a priority order — accurate → readable/attractive → efficient (elapsed time, failed repairs) — for judging any change to Archify.
Runs layout, routing, diagnostics, authoring-instruction and acceptance-rule work through a six-step loop: bound the question, choose the experiment, attribute the cause, change the smallest responsible scope, verify the claim, conclude and consolidate.
Warns that a passing check does not prove a good diagram, and allows unsuitable rules to have their scope, threshold or severity revised with before/after evidence.
Maps evidence needs to concrete commands: replay.mjs (fixed-input comparison), round.mjs (author tasks), collect.mjs (trace collection), shots.mjs (screenshot diffs).
Keeps journals and benchmarks in $ARCHIFY_TUNING_HOME outside the repo so private source content is never committed.

Who it's for

Archify maintainers and contributors working inside that repository.
Engineers tuning layout/routing quality in diagram or graph renderers.
Anyone studying rigorous evaluation methodology for agent-authored artifacts.

Examples

A bend-count rule rejects otherwise good diagrams: confirm whether the user harm is real, then adjust the threshold and keep the original failure plus before/after evidence visible.
After a router change, run replay.mjs --base <base>/archify, then classify every flipped pass/fail as a real regression, a corrected false failure, or an intended rule change.
Run round.mjs on a fresh project to measure first-draft quality and time to first usable artifact — a run that failed early is not a speed win.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/tt-a1i/archify/HEAD/.agents/skills/archify-tuning/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)

Download the ZIP with the button below.
In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.

↓ Download ZIP

Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on AI Skill Mart.
Copy the .agents/skills/archify-tuning folder from the GitHub repo tt-a1i/archify into my ~/.claude/skills/archify-tuning/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/tt-a1i/archify.git && mkdir -p ~/.claude/skills && cp -r archify/.agents/skills/archify-tuning ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

This skill is marked internal to the Archify repo, so clone it first: git clone https://github.com/tt-a1i/archify.git
Enter the repository root: cd archify
Working inside the repo needs no extra install — launch Claude Code from the root so .agents/skills/archify-tuning is discoverable.
For global use run mkdir -p ~/.claude/skills && cp -r .agents/skills/archify-tuning ~/.claude/skills/, but note the scripts/ and ../../../CONTRIBUTING.md references will break.
Prerequisites: Node 22.13 or later and Chrome/Chromium (override discovery with ARCHIFY_CHROME).
Set the journal location: export ARCHIFY_TUNING_HOME=~/.local/share/archify-tuning and never commit that directory.
Invoke it with a prompt such as "use the archify-tuning skill to design an experiment for this routing change".

View source on GitHub ↗License: Apache-2.0