Guide · Agent skill
How to use the Archify Diagram Tuning agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The archify-tuning skill is a maintainer workflow for Archify, a tool where an AI agent turns a described system into a diagram. Once installed, it gives the agent a structured process for testing and improving how well those diagrams are generated, from accuracy and layout to authoring speed. This guide explains what the skill does, how to install it, how to trigger it, and what it needs to run safely.
Archify Diagram Tuningtt-a1i/archify · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the archify-tuning ZIP from this skill's page on the directory.
- 2Unzip into a skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill.
- 4Prepare requirements. Install Node.js 22.13 or later, have Chrome or Chromium available, and check out the Archify repository you want to test.
- 5Make a natural request. Describe the change to evaluate and the diagrams to test, in plain language matching the skill's purpose.
- 6Review the journal and results. Read the recorded hypothesis, evidence and disposition to confirm what changed and what remains uncertain.
What is an agent skill, and what does this one do?
An agent skill is a folder of instructions and supporting files that an AI agent can load when a task matches the skill's description. This one, named archify-tuning, is written for people working on the Archify project itself. It is not a general diagramming helper.
Once active, the agent follows a set process: it sets up experiment rounds, generates diagrams from fixed inputs, reads agent traces, renders screenshots, compares before and after results, and records decisions in a journal. The skill asks the agent to judge the whole diagram and the authoring experience, not just to pass automated checks. Priorities are accuracy to the source system first, then readability and visual quality, then efficiency such as completion time and failed repairs.
Who should use it and who should not
The skill fits maintainers and contributors who evaluate changes to Archify's layout, routing, validation rules or authoring instructions, and who want a repeatable replay of benchmark diagrams against two versions of Archify.
It is not the right choice if you only want a diagram of your own system. It is an internal tuning process and needs the Archify repository and its tooling.
- Good fit: maintainers evaluating layout, routing or validation changes
- Good fit: contributors running controlled comparisons of generation quality and speed
- Not a fit: users who only want a diagram of their own system
- Not a fit: anyone planning to share drafts publicly without care, since the tuning home can hold content from source projects that may be private
Install the archify-tuning skill
Download the ZIP from the skill's page on this directory and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload
Check the requirements before you start
The skill needs several things that a typical user will not have by default. Gather them before asking the agent to run an experiment.
- Node.js 22.13 or later
- Chrome or Chromium, used by shots.mjs for screenshots; set ARCHIFY_CHROME to point to it if discovery fails
- A local Archify checkout to test against
- Optional: Devin CLI history or a normalized trace export file
Trigger the skill with natural requests
You do not need special syntax. Describe the maintenance task in plain language that matches the skill's purpose: evaluating a change, running a generation experiment, or comparing two versions. The agent loads the skill when the request fits its description.
Requests work best when they name the change under test and the evidence you want. Mention the diagram type you care about, since the skill is designed to use the user's chosen diagram types.
- Compare the current layout change against the base version using the benchmark diagrams.
- Run a generation experiment on the workflow diagram and report bends, crossings and time to completion.
- Evaluate whether the new authoring instructions help agents finish diagrams with fewer repairs.
Understand the tuning workflow
The skill runs a loop: bound the question, choose the experiment, attribute causes from the trace and diagram, change the smallest responsible scope, verify the claim, then conclude and consolidate. Each round records a hypothesis, the evidence that would support or refute it, and a time or attempt budget in the journal.
Supporting files cover experiment design and trace signals. Tools include collect.mjs for gathering traces, replay.mjs for fixed-input regression runs, round.mjs for creating rounds, shots.mjs for screenshots, and lib.mjs as shared code.
- Tuning home: $ARCHIFY_TUNING_HOME, defaulting to ~/.local/share/archify-tuning, holding journal.md, bench/, rounds/ and replays/
- Trace input: Devin CLI history or a normalized trace export from another agent
- Regression: replay.mjs runs fixed inputs and reports pass/fail and changed output
ARCHIFY_TUNING_HOMEWhat people use it for
- Replaying benchmark diagrams against two Archify versions to see which changes altered the output
- Measuring whether new authoring instructions reduce failed repairs and time to a usable diagram
- Investigating a layout or routing problem by reading an agent trace alongside the rendered diagram
- Checking whether a validation rule is unsuitable by comparing failing checks with the actual diagram
Frequently asked questions
Is the archify-tuning skill free to use?
The skill is published under the Apache-2.0 licence, as listed on this page. Check the licence file in the Archify repository for the full terms before redistributing.
What do I need to install archify-tuning?
Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps. You also need Node.js 22.13 or later, and a local Archify checkout to test against.
Does the skill run scripts and use the network?
Yes. The skill runs scripts and uses the network, so review the scripts folder before installing and run it only against repositories you trust. Keep the tuning home outside the repository, because it can contain content from source projects.
Can I use archify-tuning to make a diagram of my own system?
Not as intended. The skill is an internal tuning process for maintainers and contributors and needs the Archify repository and its tooling. If you only want a diagram, use Archify directly rather than this skill.
What is ARCHIFY_CHROME for?
shots.mjs needs Chrome or Chromium to render screenshots. If the skill cannot find your browser, set ARCHIFY_CHROME to the path of the executable.
What is the difference between this skill and a normal Archify run?
A normal run produces one diagram. This skill compares versions and experiments, uses fixed inputs for replays, and records decisions in a journal, so you can judge whether a change really improves accuracy, readability and speed.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use archify-tuning if you maintain or contribute to Archify and need repeatable, evidence-based comparisons of layout, routing, validation or authoring changes. If you only want a diagram of your own system, or you cannot provide the Archify repository, Node.js and Chrome, pick a general diagramming workflow instead.
Download the skill