Add Power Platform Requirements Architect skill - #259
Conversation
There was a problem hiding this comment.
Pull request overview
Adds a new Copilot Studio skill submission under submissions/power-platform-requirements-architect/ that guides a structured Power Platform requirements/architecture interview and produces a single canonical YAML handoff (POWER_PLATFORM_SOLUTION_SPEC_V1) for a separate coding agent.
Changes:
- Introduces the skill definition (
SKILL.md) with workflow phases, output/serialization rules, and completion gates. - Adds a YAML handoff template (
assets/solution-spec.template.yaml) defining the canonical schema and authority/stop conditions. - Adds supporting reference materials covering interview strategy, discovery coverage, architecture decisions, surface-specific specification guidance, handoff contract, and a manual self-check.
Reviewed changes
Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| submissions/power-platform-requirements-architect/SKILL.md | Skill runtime instructions, scope boundaries, workflow phases, and output rules. |
| submissions/power-platform-requirements-architect/README.md | Human-facing overview, usage guidance, and Copilot Studio adoption instructions. |
| submissions/power-platform-requirements-architect/metadata.json | Submission metadata (name/description/platform/tags/author/version/dates). |
| submissions/power-platform-requirements-architect/assets/solution-spec.template.yaml | Canonical YAML schema template for the single handoff artifact. |
| submissions/power-platform-requirements-architect/references/handoff-contract.md | Defines the final response shape and the coding-agent execution contract. |
| submissions/power-platform-requirements-architect/references/self-check.md | Manual checklist and pass/block outcomes before emitting the YAML. |
| submissions/power-platform-requirements-architect/references/interview-strategy.md | Core elicitation approach and sequencing guidance. |
| submissions/power-platform-requirements-architect/references/discovery-checklist.md | Domain coverage checklist to ensure requirements completeness. |
| submissions/power-platform-requirements-architect/references/architecture-decisions.md | Structured prompts for key architecture decisions (AD-*). |
| submissions/power-platform-requirements-architect/references/dataverse-spec.md | Guidance for specifying data layer details precisely (tables/columns/choices/etc.). |
| submissions/power-platform-requirements-architect/references/automation-spec.md | Guidance for specifying Power Automate flows (triggers, identity, error handling, idempotency). |
| submissions/power-platform-requirements-architect/references/app-and-agent-spec.md | Guidance for specifying app/agent surfaces (screens, states, delegation, accessibility). |
| submissions/power-platform-requirements-architect/references/coding-agent-handoff.md | Rules for decomposing into traceable work items and dependency-safe build order. |
|
@microsoft-github-policy-service agree |
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 13 out of 13 changed files in this pull request and generated no new comments.
Suppressed comments (4)
submissions/power-platform-requirements-architect/references/handoff-contract.md:50
- The manual self-check outcomes in this header don’t match the canonical
handoff.manual_self_check.statusvalues defined inreferences/self-check.mdand the YAML template, which can lead to emitting a header status that isn’t representable/traceable in the YAML.
Manual self-check: passed | passed with listed noncritical assumptions
submissions/power-platform-requirements-architect/assets/solution-spec.template.yaml:33
meta.statusin the YAML template includessuperseded, but the status model inreferences/handoff-contract.mddefines onlydraft,in_review, andapproved. This inconsistency can lead to invalid/undefined status values in generated handoffs.
meta:
spec_version: "1.0"
status: draft # draft | in_review | approved | superseded
tier: M # S | M | L
revision: 1
submissions/power-platform-requirements-architect/assets/solution-spec.template.yaml:207
- The template uses
SET_AT_DEPLOYfor an unresolved environment-variable value, but the handoff contract’s placeholder rules only defineSET_AT_BUILD(and require it map to aDEP-entry). Using an undocumented placeholder token makes the handoff rules ambiguous for the agent and any downstream tooling.
values:
dev: ""
test: ""
prod: "SET_AT_DEPLOY"
used_by: []
submissions/power-platform-requirements-architect/references/automation-spec.md:68
- This reference instructs authors to use
SET_AT_DEPLOY, but the canonical placeholder rules inreferences/handoff-contract.mdonly defineSET_AT_BUILD. Aligning terminology avoids producing specs that violate the contract.
- Every connector use becomes a **connection reference** (`CR-`), with connector, tier, auth model, consenting identity, and whether consent is a human gate (it almost always is).
- Every environment-specific value becomes an **environment variable** (`EV-`): site URLs, list names, group ids, mailboxes, API base URLs, thresholds someone will want to change without editing the flow, feature toggles.
- Secrets go in an Azure Key Vault-backed environment variable, never in a flow, never in a spec file. If the spec would otherwise contain a credential, write `SET_AT_DEPLOY` and name the owner.
|
Where is the generated YAML supposed to go? We’re not aware of a Power Platform coding agent that consumes |
Adds Power Platform Requirements Architect, a Copilot Studio skill that turns a Power Platform solution idea into a complete YAML requirements handoff for a coding agent.
The skill:
conducts structured stakeholder discovery;
defines requirements, architecture decisions, security, data, integrations, and acceptance criteria;
creates traceable work items, dependencies, deployment inputs, and build order;
separates requirements approval from implementation readiness.
This skill performs requirements engineering only. It contains no scripts, external dependencies, secrets, or organization-specific information.