Classifies requirements using BABOK's four types (business, stakeholder, functional, non-functional, transition), checks each against the IEEE 830 quality bar (correct, unambiguous, verifiable), and separates in-scope from explicitly out-of-scope items. Use before requirements are locked in, when scope keeps drifting or vague requirements like \"the system should be fast\" are passing unchallenged.
Installs into .claude/skills of the current project.
Are you the author of Requirements And Scope Framing?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/pilot2service-requirements-and-scope-framing)
---
name: requirements-and-scope-framing
description: "Classifies requirements using BABOK's four types (business, stakeholder, functional, non-functional, transition), checks each against the IEEE 830 quality bar (correct, unambiguous, verifiable), and separates in-scope from explicitly out-of-scope items. Use before requirements are locked in, when scope keeps drifting or vague requirements like \"the system should be fast\" are passing unchallenged."
---
# Requirements & Scope Framing
## Purpose
Frames the problem and requirements into a clear, testable whole.
## Anchored in research
- BABOK — requirements
- IEEE 830
## Method
1. **Capture the raw problem statement and classify each candidate
requirement** using BABOK's four requirement types — business requirements
(the why: goals and objectives), stakeholder requirements (the needs of a
specific stakeholder group), solution requirements split into functional
(what the solution must do) and non-functional (quality attributes:
performance, security, usability), and transition requirements (what's
only needed to move from the current state to the future one). Anything
that doesn't fit cleanly gets flagged for further elicitation before
scoping continues.
2. **Elicit missing requirement types with structured techniques** (BABOK:
interviews, workshops, document analysis, observation) rather than
inferring them silently — any requirement not sourced from a stakeholder
or document is an assumption and must be marked as one.
3. **Check each requirement against the IEEE 830 SRS quality checklist:** is
it correct, unambiguous, and verifiable (testable), and is it consistent
with the other requirements and traceable to its source? A requirement
that fails "verifiable" (e.g. "the system should be fast") gets rewritten
with a measurable threshold, or flagged as needing one.
4. **Separate in-scope from explicitly out-of-scope items in a single
list** — an unscoped boundary is as costly as a wrong requirement, because
it seeds disagreement later.
5. **Rank requirements on IEEE 830's importance/stability dimension** — which
are core and stable versus likely to change — so that scope decisions made
under pressure protect the right things first.
6. **Validate the resulting scope statement with stakeholders explicitly**,
tracing each requirement back to whoever asked for it (traceability)
before treating the scope as locked.
## What this skill does NOT do
- Doesn't make the final decision for you — it produces a structured draft to
support a human decision.
- Doesn't confirm figures, market data, or competitor data from memory — it
uses the inputs you provide, or marks an assumption clearly
(`[assumption — verify]`).
- Doesn't finalize a requirements specification without stakeholder
confirmation.
## Refinement notes
Areas to keep deepening with real practice:
- your own rules of thumb and heuristics for this technique
- concrete templates (into [`../../references/`](../../references/))
- reference cases / your own examples
- what this skill deliberately does *not* do (guardrails, common mistakes) —
add to the list above
This is an internal working note, not a claim about the skill's current
usability. Track depth privately via the `maturity` field in
`skills_index.json` (see
[`../../../meta/maturity_levels.md`](../../../meta/maturity_levels.md)).
**Don't add new fields to the frontmatter** — `name` and `description` are
the only ones allowed (see
[`../../../meta/frontmatter_schema.md`](../../../meta/frontmatter_schema.md)).
## Continue from here
- Next in this pack: [`../stakeholder-analysis-and-raci/SKILL.md`](../stakeholder-analysis-and-raci/SKILL.md) — Maps stakeholders on a power/interest matrix and assigns responsibilities with RACI.
- A ready-made skill chain for this situation: see [`../../../playbooks/`](../../../playbooks/)
- This pack's shared guardrails: [`../../CLAUDE.md`](../../CLAUDE.md)
## References
- [`../../references/`](../../references/) — the pack's shared background material
- [`../../CLAUDE.md`](../../CLAUDE.md) — the pack's shared guardrails