| name | aidlc-requirements-review-skill |
| description | Evaluate product artifacts — requirements, user stories, personas, and wireframes — for clarity, value, testability, traceability, and scope discipline. Applies wherever product-facing work is reviewed before it moves downstream.
|
Requirements Review
Definition
Assess whether a product artifact is ready to move forward. This skill reviews the shape of product thinking: are requirements unambiguous and verifiable, do stories trace to requirements, are personas consistent across artifacts, is scope explicit, are trade-offs and assumptions surfaced. It is the product-side counterpart to architecture review and code review.
Principles
- Every requirement must be verifiable as pass/fail. Vague language ("fast," "user-friendly," "robust") is a defect, not a placeholder.
- Every story must trace back to at least one requirement. Stories without a parent requirement are scope creep.
- Every requirement must be covered by at least one story. Uncovered requirements are dropped work.
- Acceptance criteria must be specific, observable, and testable. "Works correctly" is not acceptance criteria.
- Scope boundaries must be explicit. Silence is not exclusion — out-of-scope items must be listed.
- Assumptions must be flagged as assumptions, not stated as facts.
- Personas must be consistent across requirements, stories, and wireframes. A persona introduced in stories that doesn't appear in personas.md is a gap.
- User journeys must be complete. Happy path without error paths is half-finished.
- Non-functional requirements must be measurable. "Highly available" is not measurable; "99.9% monthly uptime" is.
- Prioritization must be explicit. If everything is P0, nothing is P0.
- Edge cases and error scenarios must be addressed, not buried in happy-path stories.
- Review at the abstraction level of the artifact. Do not require implementation detail before the relevant stage introduces it.
Definitions
- Verifiable — a requirement or acceptance criterion that can be objectively checked as pass or fail
- Traceable — an artifact that can be linked back to its parent (intent → requirement → story → screen) and forward to its children
- Atomic — a requirement or story that addresses one concern; if it has "and," it usually splits
- INVEST — Independent, Negotiable, Valuable, Estimable, Small, Testable (story quality criteria)
- Coverage gap — a requirement with no corresponding story, or a story with no parent requirement
Application
When applied to review work, this skill flags:
- Requirements that are vague, untestable, or use weasel words ("should," "may," "preferably")
- Requirements without acceptance criteria or measurable thresholds
- Compound requirements that should be split (FR-X with multiple distinct capabilities joined by "and")
- Non-functional requirements without measurable targets
- Missing scope boundaries — no out-of-scope section, no assumptions section
- Assumptions presented as facts
- Stories that don't follow "As a [persona], I want [goal], so that [benefit]"
- Stories without acceptance criteria in Given/When/Then form
- Stories with acceptance criteria that are not testable
- Stories that don't trace to any requirement (orphan stories)
- Requirements with no covering story (orphan requirements)
- Personas that appear in stories but not in personas.md (or vice versa)
- Personas that are generic stereotypes instead of grounded archetypes
- Wireframes that don't match the screen flows implied by stories
- Wireframes missing screens needed by stories (or screens with no story justification)
- Missing edge cases, error states, or unhappy paths
- Prioritization that isn't explicit, or "everything is critical"
- Scope creep — content in the artifact that doesn't trace to the original intent
- Inconsistencies across artifacts (intent says X, requirements say Y, stories say Z)
When applied to producer feedback (contributor mode rather than reviewer), this skill manifests as: pointing out the same gaps but framed as "consider adding…" rather than "missing…", and offering specific suggestions rather than verdicts.
Verdict Format
When this skill is used at the review stage, return:
- READY — the artifact is fit to move to the next stage
- NOT-READY — the artifact has blocking gaps; list each finding with the artifact section, the principle violated, and a concrete suggestion to fix it
Findings should be specific (file and section reference), actionable (what would resolve it), and proportional (do not block on stylistic preferences).
