Scope Bounder

Scope Bounder — The Boundary Surfacer

A thin coordinator that calls assumption-excavator (philosophy/logic) with depth=surface, category=scope and translates the output into the scope axis on statement-grader.

The hardest part of scope-bounding is recognizing that universal statements are usually wrong; the easiest part is asking "for whom, when, and in what context?" This skill makes the easy part routine.

The Three Boundary Types

| Boundary | What it specifies | Failure signal | |---|---|---| | Time | Period the statement applies to | "Always" implicitly — no temporal anchor | | Population | People, accounts, units the statement applies to | "Customers" with no segmentation; "users" without cohort | | Context | Conditions, environment, regime under which the statement holds | No mention of regime, market state, or operational context |

A statement that explicitly bounds all three is well-scoped. A statement that bounds two of three is acceptable. One or zero boundaries explicit triggers the re-state loop on this axis.

Process

Step 1 — Call assumption-excavator

Pass:

• The statement
• Domain context
• depth=surface (don't excavate every assumption — only scope)
• category=scope

Receive:

• Hidden scope assumptions across the three boundary types
• For each: visibility (hidden/semi-visible/visible-but-unchallenged), contestability, load-bearing assessment

Step 2 — Categorize each boundary

For each of (Time, Population, Context):

| Status | Meaning | |---|---| | Explicit | The statement names this boundary directly | | Implicit-defensible | The boundary is implicit but clearly inferable from context | | Implicit-questionable | The boundary is implicit and ambiguous; a reasonable reader could interpret differently | | Missing | The statement makes no claim about this boundary; it's universal-by-default |

Step 3 — Score the scope axis

| Pattern | Scope axis score | |---|---| | All three explicit; defensible | 5 | | Two explicit; one implicit-defensible | 4 | | One explicit; two implicit-defensible | 3 | | One or two explicit; one or more implicit-questionable | 2 | | One or zero explicit; rest missing or questionable | 1 |

Step 4 — Generate hints

For each Missing or Implicit-questionable boundary, produce a hint:

• "Specify the time period: is this Q3 2026? The next 12 months? Indefinitely?"
• "Specify the population: enterprise customers? Free-tier users? All segments?"
• "Specify the context: under current pricing? After the rebrand? In the EU regulatory regime?"

Output Format

SCOPE — [first 60 chars of statement...]
─────────────────────────────────────────────
Time:       [explicit | implicit-defensible | implicit-questionable | missing] | [signal]
Population: [status] | [signal]
Context:    [status] | [signal]

Score: [1-5]
Hints (per missing/questionable boundary):
  - [hint]

Output Contract for six-eyes

Feeds the scope axis on statement-grader (axis 3). Hints feed the re-state loop's input to Phase 4.

Scope Boundaries (of the skill itself)

• scope-bounder handles: the three-boundary check and scoring.
• scope-bounder does NOT: evaluate whether the bounded scope is correct — it evaluates whether the scope is explicit. A confidently-bounded but wrong scope still scores well; that's a different audit (often answerability).

Failure Modes

| Failure | Response | |---|---| | assumption-excavator unavailable | Inline a degraded version: ask the three boundary questions directly without the full assumption framework | | Statement is so short there's no scope claim to surface | Return "ungraded" and route back to Phase 1 (Intake) for restatement | | Two boundaries are tangled (population implicitly bounds time) | Note the entanglement; score the more-explicit of the two normally and flag the tangle |

Connections

• statement-grader (binding-vow) — feeds the scope axis
• assumption-excavator (philosophy/logic) — primary cross-domain call (with category=scope)
• wicked-vs-tame (binding-vow) — wicked problems often have inherently fuzzy scope; that's a property of the problem type, not a defect

Sources

• The three-boundary frame (time / population / context) is canonical in epidemiology and program evaluation; see Rothman's Epidemiology for the reference treatment.
• See [[kahneman-framing]] for the universal-by-default cognitive bias that scope-bounder targets.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: npbuilds
• Source: npbuilds/skill-library
• License: MIT
• Homepage: https://skill-library-prod.web.app/

Install and usage instructions live in the source repository linked above.