Use the runner when the AI is allowed to work beside a real Gateway.
This is the live-workflow skill. It teaches the agent how to use the Web Dev runner for discovery, dry runs, apply steps, backups, log reads, and page validation.
Why it was made
- Gives the agent discovery before writes, so it does not invent routes, tag providers, styles, or Named Queries.
- Separates dry-run, apply, validate, and rollback instead of mixing them into one risky step.
- Turns Gateway logs and readbacks into part of the build loop.
Common use cases
- Apply and validate a generated Perspective page on a reachable Gateway.
- Discover tag paths, style classes, views, routes, and Named Queries before building.
- Iterate on a page until route mapping, dependencies, and logs are clean.
- Use backups and rollback while testing AI-generated resources.
Proof style
Validate the artifact, not the promise.
The skill treats live Gateway evidence as the source of truth and avoids claiming success from generated files alone.
Evidence
Runner workflow proof
The graphic is AI-generated for the site. The checks below come from a sanitized read-only runner pass that proved the host-side skill can negotiate the API, inspect project state, validate route wiring, and check recent logs. Customer Gateways still need their own endpoint, token, security review, and approval before any write action.
- health returned the runner contract, feature list, supported actions, and mutation locks.
- gatewayInfo, routesList, and viewsList were callable as read-only discovery steps.
- pageValidate confirmed route-to-view structure without claiming browser rendering.
- logQuery used a bounded WARN/ERROR Perspective filter and returned a clean recent window.
- Write actions stay gated behind dry runs, mutation locks, explicit confirmations, backups, and rollback.
Example prompt
Use the Ignition Web Dev runner on the approved development Gateway to build and verify a Perspective downtime dashboard for Packaging Line 3.
Start with discovery before any write:
- Call health and confirm runnerVersion, stackVersion, supportedActions, features, and mutationLockedActions.
- Inspect the target project with routesList, viewsList, styleResourcesList, seedViewsList, componentSeedRead, tagProviders, tagBrowse, namedQueriesList, and gatewayInfo.
- Use discovered styles, seed components, tag folders, UDT structure, and Named Query contracts instead of guessing.
Build or update this route only:
- Route: /enterprise/acme/austin/packaging/line-3/downtime
- View: Line3/DowntimeDashboard
- Data scope: [Sample_Tags]Enterprise/ACME/Austin/Packaging/Line3
The page should show line status, active downtime reason, last stop time, runtime minutes today, a downtime Pareto chart, and a recent events table. Use parameters or custom properties for equipment paths so the view can be reused for another line.
Safety workflow:
1. Run dryRun first and report route conflicts, overwritten resources, dependency gaps, and planned backups.
2. Apply only if dryRun is clean and only with the exact confirmation string the runner requires.
3. After apply, validate with pageValidate, routesList, viewsList, viewRead, tagRead, namedQueryPreview or historyProbe where relevant, and logQuery for recent WARN/ERROR entries.
4. If browser proof is available, open the Perspective route and check that the screen renders without binding or layout errors.
5. Return the changed resources, validation evidence, backup id, rollback plan, and what still needs human review.
Do not delete live resources or overwrite unrelated project files unless I explicitly ask for that exact action.This is the kind of focused instruction the skill is designed to make useful. The value is in the checks and Ignition-specific defaults the AI uses after reading the skill.
Expected output
- Starts with runner health, Gateway info, routes, views, styles, component seeds, tag/UDT structure, and query discovery.
- Builds from discovered project patterns instead of invented component JSON, paths, or Named Query contracts.
- Uses dryRun before apply, reports conflicts and backups, and requires the exact confirmation string.
- Validates with pageValidate, viewRead, logQuery, tag/query/history readbacks, and browser proof when available.
- Returns changed resources, evidence, rollback instructions, and anything still needing human review.
Related workflows
Skills that pair with this one
Ignition Perspective import zip builder
Ignition 8.1Perspective Import Zip Builder
Create reviewable Ignition 8.1 Perspective packages.
Ignition log diagnostics helper
Ignition 8.1Ignition Log Diagnostics
Turn noisy Gateway evidence into next steps.
Ignition UDT builder
Ignition 8.1Ignition UDT Builder
Model equipment once, then scale the screens.
Lifetime membership
Want access to this skill?
The one-time membership unlocks this skill, the full Ignition AI Toolkit, the Web Dev runner materials, setup docs, and versioned Ignition 8.1 and 8.3 downloads.
No subscription. Lifetime members keep portal access to the current toolkit and future skill releases, including planned Ignition 8.3-focused skills as they are added.