Can the workflow inspect a repository?

Yes, when that repository is authorized and in scope. It can ground map content in direct code/config observations, while labeling inferences and unknowns separately. That does not turn the map into a code audit or establish runtime behavior without independent evidence.

Does the output require a diagram service?

A standalone result is designed to keep topology, controls and styling inside the artifact and run offline, without network/cookie/storage dependencies. The agent must implement and test that output from your material; the package itself contains no completed offline map. An embedded version follows the existing application’s tooling.

Can it show impact or blast radius?

It may show reachability through the authored graph and immediate relationships in a node dossier. That is a property of the modeled edges, not proof of production impact, capacity or causality. Runtime-impact claims need separate analysis evidence.

Does the ZIP contain the online example?

No. It contains workflow and reference notes, not a finished demo or bundled map engine. Provide the system description, relationship model or authorized repository and the question to answer so an interactive artifact can be built. Browser verification belongs to that generated result.

Interactive System Map

Previews from the publisher

Make a system’s main path and typed relationships explorable without losing the static overview. Interactive System Map turns an approved description, relationship list or authorized repository evidence into one architecture, workflow, sequence, data-flow or lifecycle map. Search, focus, layers, zoom and a selected-node dossier support reading; claims remain tied to supplied evidence, observations, inferences or unknowns.

What you get

One generated interactive map plus its underlying typed relationship model. Nodes have stable IDs, semantic kinds, concise labels, optional groups and evidence notes; edges record source, target, direction, relationship meaning and evidence status. One declared main path anchors short side branches.
A readable static topology with meaningful labels, protocols, actions and boundaries visible before interaction. Shape, label and color work together; stronger main-path lines and conditional/asynchronous dashes are used only when their meaning is supported.
Searchable nodes, keyboard focus, semantic layers/filtering, selected-node relationship details and bounded zoom or pan with reset. A dossier can emphasize direct neighbors and authored metadata without altering the topology or hiding essential connections behind hover.
A standalone output designed to hold its topology, controls and styling offline without network, cookies, storage or a hosted diagram service. In an existing repository, the workflow follows its component/styling conventions instead of silently replacing the stack. Wide graphs stay in a labelled contained region on phones, with legible labels and deliberate panning.
A handoff separating supplied evidence, local observations, inferences and unknowns, and describing layout choices, interactions and actual browser verification. The installable ZIP contains the workflow and reference notes only: it does not include the online example, a ready-made graph renderer or a completed offline demo.

Before you choose

Check the map in two passes: first follow the main path without touching a control, then select a node and inspect its actual relationships and evidence. Exploration should clarify the same topology, not reveal a different story or suggest proven runtime effects from an authored connection.

Good fit

A team needs one shared explanation of architecture, request paths, workflows, data flow or lifecycle states.
The relevant system relationships can be grounded in a description, typed model or authorized repository evidence.
Readers need to search and inspect relationships without depending on a hosted diagram service.

Not for

A code audit, capacity plan or proven production-impact analysis without separate supporting evidence.
A map whose required nodes, labels, identities or relationships cannot be established from authorized material.
A free-form drawing canvas, live system monitor or multi-user diagram editor.

How to use

You say/interactive-system-map Map a cold-chain request path from field sensors through alert routing and storage.

Provide the system material and the primary question for the intended reader. Identify exact trust boundaries, protocols, states, main paths and exceptions. Keep the evidence boundary explicit: label each claim supplied, observed, inferred or unknown. Stop to resolve required relationships rather than inventing labels or accessing an unauthorized source.
Inspect an existing project’s dependencies, framework, components, tokens and diagram conventions before implementation; retain its idioms. Use a self-contained file for a standalone result. The package specifies no universal build/runtime dependency, so tooling follows the chosen host; it supplies authoring instructions rather than a finished application.
Choose one grammar—architecture, workflow, sequence, data flow or lifecycle—based on the question. Normalize stable node/edge data and declare the main path. Preserve meaningful edge protocols, action labels, direction and synchronous/asynchronous behavior; visual simplification must not remove the evidence.
Compose the main path first, place branches close to their owners, group related nodes and keep crossings rare. Repair layout by removing genuinely redundant edges, regrouping, spacing, repositioning labels and rerouting before shortening wording. Do not counterfeit a fit by hiding overflow or shrinking labels below readability.
Add search, focus, semantic layers, dossier, bounded zoom/pan and reset with accessible names and keyboard behavior. Use procedural SVG for topology and restrained inline SVG controls; add cleared imagery only when it improves understanding. Direct-neighbor or reachability emphasis must follow authored edges, not invent relationships or imply runtime causation.
Remove standalone network/cookie/storage dependencies, keep essential topology visible without animation, honor reduced motion and contain the graph at narrow widths. Controls should stack while the map remains legible, using deliberate horizontal panning where needed instead of compressing all nodes.
Verify every label and edge direction against evidence, then test search, focus, layers, zoom, reset, theme behavior, keyboard use, wide/narrow containment, offline opening and console output in the actual generated artifact. Deliver the typed model and evidence/layout/interaction handoff, reporting observed tests separately from untested expectations.

Requirements and costs

You bring: A system description, repository, or typed node-and-edge list · The primary question the map must answer · Trust boundaries, main paths, and exceptions that need emphasis
Dependencies: Not collected yet
Outside costs: Not collected yet
Publisher testing: v1.0.0 — Claude Code / Claude Desktop / Cowork / Codex / ChatGPT