Vulcan

Production-first, constrained multi-agent execution for repeatable, correct outcomes.

Deterministic handoffs Quality gates Evidence-driven Operational workflows

What Vulcan is

Vulcan is not a chatbot. It is a disciplined agent system designed to complete real tasks with structured inputs, explicit roles, and quality checks—so the output is usable, repeatable, and less dependent on “prompt luck.”

When to use Vulcan

When you need a correct, actionable plan (not brainstorming).
When the problem can be handled with a bounded workflow (checklists, diagnostics, steps, validation).
When you want auditable reasoning: what was assumed, what was verified, and what still needs confirmation.

What Vulcan produces

Step-by-step execution plan (commands, click paths, checklists).
Verification steps (how to prove it worked).
Failure handling (fast fallbacks and common fixes).
Definition of Done (clear stop condition).

How Vulcan works (in plain English)

Intake: capture goal, constraints, environment details.
Plan: generate a short execution sequence (no essays).
Quality gate: check for missing steps, unsafe actions, scope creep.
Execute: user runs steps; Vulcan adapts only based on observed results.
Close: confirm “done” using explicit verification criteria.

How to request work from Vulcan

Use this prompt pattern:

GOAL:
- (one sentence)

ENVIRONMENT:
- OS:
- Cloud/account:
- Constraints (prod-safe, no breaking changes):

INPUTS:
- Logs / screenshots / errors:

OUTPUT FORMAT:
- Instructional steps only
- Verification steps
- If-fails fixes

Example request

GOAL:
- Deploy two static instruction pages behind HTTPS.

ENVIRONMENT:
- Windows 10/11, gcloud installed, Google Cloud project linked to billing.
- Production-safe: add-only DNS changes, no disruption to existing products.

INPUTS:
- Domain: agilesolutionsinc.org
- Hostnames:
  - vulcan.instructions.agilesolutionsinc.org
  - olumpus.instructions.agilesolutionsinc.org

OUTPUT FORMAT:
- Copy/paste commands + verification checks.

What not to use Vulcan for

Open-ended creative ideation where “many possible answers” is the goal.
Problems that require deep ambiguity exploration (use Olumpus instead).
Tasks that require privileged or destructive actions without approvals.