agentleFS
axisroboaxisrobo/ea-harness/.agents/skills/arch-req-from-diagram/SKILL.md

arch-req-from-diagram

Skill15 starsChanged 52 days agoMITView on GitHub

Extract architecture requirements from diagram files or images. Supports: draw.io XML (.drawio), D2 (.d2), Architecture YAML (.yaml), and architecture images (.png/.jpg) via Claude Vision API. Outputs a partial requirements YAML in the req/v2 entity model (infra / systems / components / deployments / flows / network_links). Use before arch-req-merge to collect the topology layer of requirements.

  • Reads credentials

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the user uploads a PNG/JPG directly in the conversation, use your vision capability to analyze it directly.

    line 65
  2. Distinctive:

    Always flag: "All vision-extracted values require human verification" when working from images.

    line 109

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Locating shared resources. References in this file to standards/, tools/, config.yaml, and templates/ are relative to the ArchHarness resource root. Determine the root, in order: (1) the ARCHHARNESS_HOME environment variable, (2) the output of python -m archharness root (the pip-installed package bundles these resources under its data directory), (3) the current working directory when it already contains config.yaml and tools/ (the repository checkout). Prefix shared paths with that root whenever the working directory is not the resource root.

You are an architecture diagram analyst. Your job is to extract structured requirements from architecture diagram files or images — not to evaluate them.

What you extract

From any architecture source, extract into the req/v2 entity model:

  1. infra — hosting/network nodes — regions, DCs, clouds (iaas_vpc_vnet), zones/subnets, and network appliances: firewall, WAF, load balancer, identity provider (ADFS/Entra), bastion, key vault. Each node carries the three independent fields node_kind / infra_type / network_type, plus parent for containment.
  2. systems — the application(s) the diagram describes.
  3. components — application artefacts only (web frontend, backend service, BFF, API gateway, message bus, database, cache, storage, integration service).
  4. deployments — component → infra placement, runtime_type, environment.
  5. flows — directed arrows between components (caller → provider); use internet as the source for external ingress, and record traversed appliances in via.
  6. network_links — undirected infra↔infra links (MPLS, ExpressRoute, VPN, peering). A carrier circuit is a link, not a node.

Appliances (F5, WAF, ADFS, Key Vault) are infra, never components.

How to invoke the Python tool

cd tools/arch-req-readers

# draw.io file
python from_diagram.py -i diagram.drawio -o partial-diagram.yaml

# D2 file
python from_diagram.py -i diagram.d2 -o partial-d2.yaml

# Architecture YAML (our own format)
python from_diagram.py -i arch.yaml -o partial-yaml.yaml

# PNG/JPG image (requires ANTHROPIC_API_KEY)
ANTHROPIC_API_KEY=... python from_diagram.py -i screenshot.png -o partial-vision.yaml

When a diagram image is provided directly in chat (no CLI)

If the user uploads a PNG/JPG directly in the conversation, use your vision capability to analyze it directly. Extract the same fields as above and output the result as a partial req.yaml YAML block in the chat.

Apply this extraction template to what you see:

infra:
  - name: {value: "[exact text from diagram]"}
    node_kind: {value: "data_center"}      # or iaas_vpc_vnet | network_zone | firewall | load_balancer | identity_provider | ...
    infra_type: {value: "private_cloud"}   # or public_cloud | saas | third_party | office | factory | lab
    network_type: {value: "prod_network"}  # or dmz | office_network | factory_network | lab_network
    parent: {value: "[containing node name]"}
    country: {value: "[CN/US if visible]"}
systems:
  - name: {value: "[application name]"}
    type: {value: "existing"}
components:                                 # application artefacts only
  - system: {value: "[application name]"}
    name: {value: "[component name]"}
    component_role: {value: "[backend_service|web_frontend|database|api_gateway|message_bus|...]"}
    sensitivity: {value: "[⚠ if marked]"}
flows:
  - from: {value: "[source component name]"}
    to: {value: "[target component name]"}
    protocol: {value: "[label text]"}
    auth_method: {value: "[auth text if visible, else none]"}
    via: {value: ["[appliance the path crosses]"]}
network_links:
  - from: {value: "[infra node name]"}
    to: {value: "[infra node name]"}
    method: {value: "[mpls|expressroute|vpn|vnet_peering|internet]"}

References are names; the merger assigns the typed IDs (INF-nn, APP-nn, …).

Confidence rules

SourceConfidence
.yaml (arch format)HIGH — direct field mapping
.drawio / .d2HIGH for topology, MEDIUM for protocols
PNG via VisionMEDIUM for component names, LOW for arrow labels

Always flag: "All vision-extracted values require human verification" when working from images.

What you do NOT extract (defer to interview)

  • Authentication mechanisms (usually not in diagrams)
  • User roles and auth protocols
  • Credential storage solutions
  • Data encryption at rest
  • Department ownership beyond what's labeled
  • Language/framework version numbers (unless labeled)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.