Build an Interactive System Map from Supplied Evidence
Choose one architecture question, author a typed node-and-edge model and test an offline map with search, node details and bounded zoom. Keep inferred relationships distinct from observed behavior.
Fastest way
Ask your AI to do it
Paste this into the AI agent you already use:
Use Interactive System Map with my explicitly supplied description or permitted repository evidence. Choose one question and diagram grammar. Author stable typed nodes and edges, exact labels, source references, evidence statuses and one main path. Create an offline map, model.json and handoff.md in a fresh folder. Test search, keyboard selection, immediate neighbors, zoom bounds, reset, narrow containment and console. Keep unknowns visible; do not claim graph reachability proves runtime impact or security.
Does this skill include an automatic architecture scanner?
Interactive System Map v1.0.0 is an instruction workflow for your agent. The reviewed package contains SKILL.md and three references; it does not ship a map generator, scanner or bundled demo file. Your agent authors the map from your permitted description, code evidence or typed specification. It must not invent missing infrastructure or runtime relationships.
Choose one question and one diagram grammar
Ask a narrow question, such as how a supplied export request moves from an interface to a stored artifact. Choose architecture, workflow, sequence, data flow or lifecycle rather than mixing them into one canvas. Record the source and revision; mark every statement as supplied, observed, inferred or unknown.
Build a source inventory before explaining a repository
Author stable nodes and typed relationships first
Read references/map-authoring.md and references/icon-library.md. Use stable node IDs, semantic kinds, concise labels, groups and evidence notes. Each edge needs a source, target, direction, exact relationship label and evidence status. Declare one main path before adding exceptions.
Explore the fictional map and its visible relationship model
| ID / kind | Role | Evidence boundary |
|---|---|---|
| browser · client | Submit the export brief | Supplied fictional requirement |
| gateway · service | Check and enqueue the request | Supplied fictional requirement |
| worker · process | Create the artifact | Supplied fictional requirement |
| archive · store | Store the artifact | Retention duration is unknown |
Ask your agent to author a complete offline artifact
Install and select the skill, supply your reviewed model and ask for map-output/index.html, model.json and handoff.md in a fresh directory. These are requested output paths, not files already included in the package. For an existing application, inspect its actual framework and components before choosing an implementation. For a standalone map, inline the required CSS, JavaScript and SVG and remove remote fonts, account and storage dependencies.
After those files exist, the following standard-library server can preview your output locally. Open http://127.0.0.1:5376/ and stop it with Ctrl+C. The server does not generate a map or inspect a repository.
python3 -m http.server 5376 --bind 127.0.0.1 --directory ./map-outputTest the topology and the interactions separately
- Reconcile every node, endpoint and edge label with the model and its cited source. A declared asynchronous queue should not become an unlabeled synchronous arrow. Missing production evidence remains unknown.
- Search for an exact node label, select it with keyboard and pointer, inspect its immediate relationships and clear the search. Selection can emphasize existing neighbors but must not change or invent topology.
- Check both zoom limits, reset, theme, reduced motion and readable labels. At phone width, contain the wide map inside a labelled scroll region; do not shrink the whole diagram into illegibility or let it overflow the document.
- Inspect the browser console and required network resources. An offline claim requires the artifact to work without external scripts, fonts, images, cookies or saved account state; a static screenshot cannot establish those behaviors.
Explain what the map does and does not prove
Deliver the map, typed model, source locations and revision, assumptions, layout decisions and actual browser checks. Graph reachability describes authored connections; it does not establish runtime failure impact, capacity, security or production correctness. Update the evidence and inspect the diagram again when the real system changes.
Verify actual API behavior separately from an architecture map
Capture a reproducible defect instead of inferring impact from a line