Relay discovery
Socratic discovery for Power Platform projects. Used by Scout to turn a one-paragraph brief into a fully validated requirements document. Asks questions one at a time, proposes approaches, validates design sections before writing. Adapted from Superpowers brainstorming skill — no external dependency required.From its SKILL.md
npx -y skills add Genb2nu/Relay --skill relay-discoveryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
5.4 KB, as published. Nobody here has run it
Relay Discovery — Requirements Gathering
Adapted from Superpowers brainstorming skill for Power Platform project discovery. Scout uses this skill to turn a brief into a validated requirements document.
Core Principle
Never start writing requirements until you understand the full picture. A half-understood brief produces a half-correct solution. One wrong assumption at this stage costs 10x to fix in Phase 5.
Anti-Pattern: "The Brief Is Clear Enough"
Every project needs discovery — even ones that seem obvious. "Build a request-management app" could mean:
- Canvas App only, or Canvas + Model-Driven
- Simple approval or multi-level escalation
- Employee self-service or HR-managed
- Dataverse or SharePoint backend
The design can be short for simple projects, but you MUST validate before writing requirements.
Discovery Process (Scout follows this exactly)
Step 1 — Read existing context first
Before asking a single question:
- Check if
context/folder exists and read all files in it - Check if
.relay/context-summary.mdexists (from/relay:load) - Read any uploaded wireframes, BRDs, task lists
- What do you already know? What gaps remain?
Step 2 — Ask questions ONE AT A TIME
Never ask more than one question per message. If a topic needs more exploration, come back to it after the user answers.
Prefer multiple choice over open-ended:
❌ "Who will use this system?"
✅ "Who are the primary users? A) Only employees submitting requests,
B) Employees + managers who approve, C) Employees + managers + HR admins,
D) Something else"
Question order for Power Platform projects:
- Personas — who uses it and what do they do?
- Core workflow — what is the main process end-to-end?
- Data — what entities does this track?
- Security — what data should each persona NOT see?
- Integration — SharePoint? Teams? Outlook? External systems?
- Existing solutions — is there anything already built we need to work with?
- Out-of-scope — what should this NOT do?
Stop asking when you can answer all of these:
- Who are all the personas and their access boundaries?
- What is the complete workflow from start to finish?
- What data needs to be stored?
- What should each persona NOT be able to see or do?
Step 3 — Propose 2-3 approaches before committing
Once you understand the requirements, propose options:
I see two main approaches for this:
Option A: Canvas App + Dataverse
- Employee submits via Canvas App, manager approves via Model-Driven App
- Full control over UI, offline capable
- More build effort
Option B: Model-Driven App only
- Standard Dynamics-style forms for all users
- Faster to build, less custom UI
- Less polished employee experience
I recommend Option A because [reason].
Which direction fits better?
Step 4 — Present design in sections, validate each one
Don't dump the full requirements at once. Present section by section:
Section 1 — Personas and access boundaries
[present, ask "does this look right?"]
Section 2 — Core workflow and user stories
[present, ask "anything missing?"]
Section 3 — Data model (entity names and purpose — no columns)
[present, ask "are all entities covered?"]
Section 4 — Out of scope and open questions
[present, ask "any gaps?"]
Only proceed to writing the requirements document after all sections are approved.
Scope Check — Multi-Subsystem Projects
Before detailed questions, assess scope. If the brief describes multiple independent subsystems (e.g. "HR platform with leave, expenses, performance, and training"), flag it immediately:
"This sounds like 4 separate projects. Each should have its own requirements → plan → build cycle. Which one should we start with?"
Don't spend questions refining a project that should be decomposed first.
Column Definitions — NOT Scout's Job
Scout captures entity NAMES and PURPOSE only. Scout does NOT define:
- Column names or data types
- Relationships between tables
- Choice set values
- Required vs optional fields
That is Drafter's job. If you find yourself writing column lists, stop.
What Goes in requirements.md
After approval, write docs/requirements.md:
# Requirements — <Project Name>
## Personas
### <Name>
- Description
- Key actions
- Access boundaries (what they CANNOT see/do)
## User Stories
### <Persona>
- As a <persona>, I want to <action> so that <value>
[minimum 5 per persona]
## Entities (names and purpose only)
- <EntityName>: <what it represents, one sentence>
## Apps
- <Canvas App name>: <personas and purpose>
- <Model-Driven App name>: <personas and purpose>
## Integrations
- <System>: <what data flows where>
## Out of Scope
- <item>
## Open Questions
- <question that needs a decision before or during planning>
Handoff to Drafter
After writing requirements.md, tell Conductor:
Requirements complete. Personas: <N>. User stories: <N>. Entities: <N>. Open questions: <N>.
Ready for Phase 2 — Planning (Drafter).