agentleFS
stacklokstacklok/toolhive-studio/.cursor/skills/deep-links/SKILL.md

deep-links

Skill173 starsChanged 5 months agoApache-2.0View on GitHub

Deep links in ToolHive Studio. Use when implementing, debugging, or asking about deep link features (toolhive-gui:// protocol), adding new deep link intents, understanding the deep link architecture, IPC model, or platform/packaging support.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    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:

    All changes happen in common/deep-links.ts:

    line 66

≈ “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

Deep Links in ToolHive Studio

Deep links allow external systems (browsers, terminals, other apps) to trigger navigation inside ToolHive Desktop via the toolhive-gui:// custom protocol.

About this document: Much of the content in the reference docs is the result of research into how other Electron apps implement deep links. Some design decisions are implemented; others describe the intended direction but are not yet in the codebase. The base skill reflects the current implementation. The reference docs reflect the research and design intent — read them with that in mind, and update them when relevant implementation decisions change.


URL Schema

toolhive-gui://v1/<intent>[?<query>]

Examples:

  • toolhive-gui://v1/open-registry-server-detail?serverName=fetch — open a registry server detail page
  • toolhive-gui://v1/open-registry-server-install?serverName=fetch — open the registry server detail page and auto-open the install dialog
  • toolhive-gui://v1/open-registry-skill-detail?namespace=io.github.stacklok&skillName=skill-creator — open a registry skill detail page
  • toolhive-gui://v1/open-registry-skill-install?namespace=io.github.stacklok&skillName=skill-creator&version=v1.0.0 — open the skill detail page and auto-open the install dialog with the reference (and optional ?version tag) prefilled. Tag-only — OCI digests (sha256:…) are intentionally not supported because safeIdentifier rejects colons; users wanting digest pinning can paste it into the dialog directly.

The v1 segment is the version. The intent is a kebab-case action name. Query params carry intent-specific data.


Current Implementation

Key Files

FileRole
common/deep-links.tsSingle source of truth. All deep link definitions: intent name, Zod param schema, navigation target.
main/src/deep-links/parse.tsParses and validates a raw URL string using the schemas from common/deep-links.ts.
main/src/deep-links/index.tsEntry point: extracts URL from argv (Windows/Linux), waits for window ready, dispatches via IPC.
main/src/deep-links/squirrel.tsSquirrel.Windows-specific protocol registration.

IPC Channel

deep-link-navigation — sent main → renderer as a NavigateTarget ({ to: string; params?: Record<string, string> }).

The renderer receives this and calls the TanStack Router navigate() directly.

How It Works (Current)

  1. Protocol registration: On app start, app.setAsDefaultProtocolClient('toolhive-gui') registers the protocol. On Windows with Squirrel, registerProtocolWithSquirrel() is called instead (see squirrel.ts).
  2. URL extraction: On Windows/Linux, the URL arrives in process.argv. extractDeepLinkFromArgs() scans for the first toolhive-gui:// argument (safe against argv injection — see patterns doc).
  3. Parse + validate: parseDeepLinkUrl() parses the URL and runs it through the Zod discriminated union schema defined in common/deep-links.ts. Invalid links resolve to showNotFound.
  4. Window ready: waitForMainWindowReady() polls until the window is visible and not loading before dispatching.
  5. Dispatch: resolveDeepLinkTarget() converts the validated intent to a NavigateTarget, which is sent to the renderer via the deep-link-navigation IPC channel.
  6. Renderer: The renderer listens for deep-link-navigation and calls navigate(target).

Current Limitations vs. Design Intent

The current implementation only supports read (navigate) operations. The design doc proposes a confirmation flow for write/destructive operations (C/U/D), but this is not yet implemented. The IPC sends a pre-resolved NavigateTarget rather than a raw parsed intent — this simplified the initial implementation. See design doc for the full intended model.


How to Add a New Deep Link

All changes happen in common/deep-links.ts:

// 1. Define the new intent using v1DeepLink()
export const myNewIntent = v1DeepLink({
  intent: 'my-new-intent', // kebab-case, matches URL path segment
  params: z.object({
    someParam: safeIdentifier, // use safeIdentifier for user-supplied strings
  }),
  navigate: (params) => ({
    to: '/some-route/$id', // TanStack Router route
    params: { id: params.someParam },
  }),
})

// 2. Add to allDeepLinks array
const allDeepLinks = [
  openRegistryServerDetail,
  showNotFound,
  myNewIntent,
] as const

// 3. Add to deepLinkSchema discriminated union
export const deepLinkSchema = z.discriminatedUnion('intent', [
  openRegistryServerDetail.schema,
  showNotFound.schema,
  myNewIntent.schema, // ← add here
])

Test manually:

./node_modules/.bin/electron . "toolhive-gui://v1/my-new-intent?someParam=value"

safeIdentifier is defined as z.string().regex(/^[a-zA-Z0-9_.-]+$/) — use it for any param that could be user-supplied to prevent injection.


Reference Documents

For deeper background, see:

  • OS & Packaging Support — Platform-specific registration requirements (Windows, Linux, macOS) and packaging format considerations (Squirrel, Flatpak, .deb, .rpm, .dmg, AppImage, MSIX, etc.). Largely research/prior art.
  • Observed Patterns — Patterns from VS Code, GitHub Desktop, Mattermost, Element, and others: URL sanitization, argv injection, security confirmations, waiting-for-readiness patterns, telemetry.
  • Design & Decisions — Full design rationale, IPC model, error handling strategy, queue management, testing approach, and the decisions log. Some sections describe planned future behaviour not yet implemented.

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.