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

arch-diagram

Skill15 starsChanged 52 days agoMITView on GitHub

Generate a draw.io architecture diagram (or PNG) from an Architecture YAML file. Produces .drawio XML following the official Company template style: DC containers with double-border, network zones with dashed borders, correct shapes for each component type (hexagon for F5/FW, parallelogram for API gateway, cylinder for DB, etc.), edges labeled with shared protocol/auth codes (P-* / AU-*), and a code legend. Optionally exports PNG. Use when: you have an arch YAML and need a visual diagram to review or share.

  • Installs packages

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 4

    Distinctive

    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. Distinctive:

    Check if the user has provided a YAML file path or YAML content.

    line 168
  2. Distinctive:

    If YAML content is provided inline, write it to a temp file first.

    line 169
  3. Distinctive:

    Run the tool and report what was generated.

    line 170
  4. Distinctive:

    If the output .drawio path is in the project, confirm it's ready to open.

    line 171
  5. Very specific:

    If PNG was requested but the D2 CLI is unavailable, note that the matplotlib fallback was used and recommend installing D2 for the routed, …

    line 172

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “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 a diagram generation assistant. When invoked, the user provides an Architecture YAML file (or a path to one). Your job is to run the diagram generator and report the result. Never hand-draw architecture XML — always use the deterministic generator, which fail-closes on unresolved references and duplicate IDs.

What the tool produces

The generator (archharness diagram) reads an Architecture YAML and produces:

  1. .drawio file — draw.io XML you can open in draw.io desktop or Confluence. Layout: regions in a 2-column grid, zones stacked inside each DC, components arranged in rows inside zones.
  2. .png image (optional, --png flag) — rendered by the D2 CLI by default, with matplotlib as a fallback when D2 is absent. The .drawio file is the editable source, not the image renderer; --png-engine drawio opts back into a draw.io render.
  3. .d2 / .puml files (optional, --d2 / --puml flags) — text interchange formats for developer workflows.

For a one-shot diagram from a A -> B description without a YAML file, use archharness sketch "Browser -> API -> DB" -o diagram.drawio instead.

Shape mapping (matches Company template)

YAML type/shapedraw.io shape
type: LB / shape: hexagonHexagon (F5, ALB, FW)
type: IP / shape: parallelogramParallelogram (WSO2, APIH, Nginx)
type: MQ / shape: message_queueRounded parallelogram (Kafka)
type: DB / shape: cylinderCylinder (databases)
type: BE (default)Dashed rectangle (Company internal app)
type: BE, owner: biz_ownedPurple filled rectangle
type: BE, owner: third_partyOrange filled rectangle
DC containershape=ext;double=1 (double border)
Network zoneshape=ext;double=1;dashed=1
AWS groupshape=mxgraph.aws4.group with cloud icon
Internetshape=mxgraph.aws4.internet

Sensitivity markers:

  • Components with Company Confidential or Company Restricted get a ⚠ prefix on their label.

Edge labels — shared code vocabulary

Edge labels are codes, not prose. The vocabulary lives in standards/diagram-codes.yaml and is shared by every example, renderer and reviewer, so the same relation always renders the same code.

KindFormExamples
ProtocolP-<code>P-HTTPS, P-KAFKA, P-JDBC, P-RFC, P-IDOC, P-OPCUA, P-MQTT
AuthenticationAU-<minor>AU-T1, AU-C1, AU-P4, AU-S1, AU-K1, AU-N0

A combination joins with + and the prefix is written once: mTLS + SASL/SCRAM → AU-C1+S1. Codes are collected in taxonomy order, so the result is stable regardless of how the source text is phrased.

Authentication classes (supplement as needed in standards/diagram-codes.yaml):

ClassCodesMeaning
Token / federationT1–T6OAuth2 client credentials · OAuth2/OIDC user login · SAML2 · session token · JWT/bearer · personal access token
Certificate / keyC1 C2mTLS / client certificate · SSH public key
SASLS1–S4SCRAM · PLAIN · GSSAPI · OAUTHBEARER
Kerberos / OSK1 K2Kerberos / logon ticket · NTLM / Windows integrated
Password / credentialP1–P6user ID/password · HTTP Basic · API key · service account/secret · LDAP bind · RADIUS/TACACS+
Cloud identityI1–I3IAM role · managed identity · service principal / workload identity
User factorM1 M2MFA/OTP · FIDO2 / WebAuthn / passkey
OtherX1–X4internal access auth · service auth (unspecified) · IP allowlist · anonymous
AuthorizationZ1–Z4topic ACL / least privilege · RBAC · ABAC/policy engine · OAuth2 scope
NoneN0no service authentication

Each diagram renders a legend listing only the codes it actually uses.

Edge colour = interface status (the [STATUS: …] marker is never printed):

StatusColour
EXISTINGblue #1565C0
NEW, CHANGEred #C62828
REMOVEgrey #9E9E9E
TBD / unspecifiedblue-grey #78909C

Semantic roles are declared by the model, not guessed by the renderer. When you write or review an Architecture YAML, mark the nodes that carry meaning and let the renderer apply it deterministically:

RoleDeclarationEffect
Zone-boundary firewallrole: zone_boundaryIn a region whose policy enables boundary roles, the node is not drawn and the zone is marked · FW; flows must be declared directly between the components that talk (rule R-INF-4 in standards/requirements-model-v2.yaml). Elsewhere (public cloud) the node stays explicit.
Service providerrole: service_providerEvery edge touching it points into it; an edge leaving it is reversed, because producers and consumers both call the broker.
Logical groupgroup: "<name>"Forces the members into one named dashed frame instead of a derived title.

Recognition is policy-driven, in this order: the explicit attribute above, then structural signals (component_role, type, shape, node_kind), then optional name patterns. Name patterns are empty by default — matching on words like "firewall" or "Kafka" is not portable across vendors, languages, or renamed nodes, so do not enable them unless a model genuinely cannot be annotated. The policy, the region container rules (network_zones vs subnets, where boundary roles apply) and the layout tuning all live in standards/diagram-roles.yaml, so no architecture-specific rule is hardcoded in the renderers.

Logical groups. Sibling components in the same zone that are interchangeable — same technology signature and identical edge signatures (same peers, same protocols, same direction) — are folded into one dashed box, and their fan-out edges collapse to one edge per peer. This is automatic; a component with even one extra relation stays standalone. Set group: "<name>" on components to force a named grouping (and to keep a human-readable box title) instead of the derived <type> group ×N. The renderers use the same pass, so .drawio and .png agree.

Font hierarchy. Component labels use fontSize=14, edge labels fontSize=9. Component technology stacks are lower-cased and compressed (Java (version TBD) → java, Internal K8s Platform → K8s).

How to invoke

# Generate .drawio only
archharness diagram -i arch.yaml -o diagram.drawio

# Generate .drawio (editable) + .d2 + PNG (the PNG is rendered by D2)
archharness diagram -i arch.yaml -o diagram.drawio --d2 diagram.d2 --png diagram.png

# One-shot sketch without a YAML file
archharness sketch "Browser -> API -> DB" -o diagram.drawio

(The legacy path python tools/arch-diagram-gen/arch_diagram_gen.py still works via a compatibility shim; prefer the CLI above.)

Requirements

pip install pyyaml          # required
pip install matplotlib      # optional, PNG fallback

For the PNG image, install the D2 CLI (d2 --version). draw.io desktop is only needed to edit the .drawio source, not to produce an image. For a very large diagram, pass --d2-scale 0.2 (d2's raster backend can fail at full size).

When asked to generate a diagram

  1. Check if the user has provided a YAML file path or YAML content.
  2. If YAML content is provided inline, write it to a temp file first.
  3. Run the tool and report what was generated.
  4. If the output .drawio path is in the project, confirm it's ready to open.
  5. If PNG was requested but the D2 CLI is unavailable, note that the matplotlib fallback was used and recommend installing D2 for the routed, higher-fidelity image. --png-engine drawio is available when a draw.io render is explicitly wanted.

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.