| name | spec-from-requirement |
| description | Use when a user has a rough feature requirement, ticket, or product note and wants a ReAuth spec created or updated under docs/specs. This skill turns ambiguous requirements into implementation-ready specs by reading the repo docs, checking for existing specs first, and using Graphify before broad architecture scanning when graphify-out/graph.json exists. |
Spec From Requirement
Overview
Turn a rough requirement into a buildable ReAuth spec. Use this before meaningful implementation work when the request changes business rules, public APIs, persistence, flows, or auth behavior.
Workflow
- Read
AGENTS.md, docs/README.md, docs/agent/07-spec-driven-delivery.md, and docs/agent/08-documentation-system.md.
- Check
docs/specs/ first to decide whether to update an existing spec or create a new one.
- If architecture or codebase context is needed and
graphify-out/graph.json exists:
- compare the commit recorded in
graphify-out/GRAPH_REPORT.md with git rev-parse HEAD
- run
graphify query, graphify explain, or graphify path before broad repo scanning
- Read only the relevant
docs/memory/ files for the affected area.
- Write or update the spec in
docs/specs/ using docs/specs/spec-template.md.
- Keep the spec status accurate:
Draft when product or scope questions are still open
Ready when the feature is specific enough to implement
Required Output
- Explicit business rules that are independently testable.
- Concrete module impact across backend, frontend, persistence, and flows when relevant.
- API and persistence changes only when the requirement actually needs them.
- Test scenarios that can be mapped directly to implementation checks.
- Clear out-of-scope boundaries.
- Open questions instead of invented product policy when intent is still unresolved.
Guardrails
- Do not implement code in this skill.
- Do not create a new spec when an existing spec should clearly be updated.
- For auth and security features, explicitly cover system prerequisites, realm policy, and flow composition.
- If the requirement is too vague to make a product decision, preserve the ambiguity in
Open Questions.