vast-transform

vast-transform

When to use

Invoke when you have an abstract or prose description — a pitch, a project brief, a backlog narrative, a strategy memo, a feature list — that is not yet in VAST shape, and you want it reshaped into a proposed Vision / Architecture / Strategy / Tactics draft. This skill is generative: it produces a VAST-shaped document from prose, extracting the latent Vision and sorting the rest into layers.

Common triggers:

• "Transform this into VAST"
• "Make this VAST-shaped" / "VAST-ify this"
• "What's the Vision / Architecture here?"
• "Put this in VAST terms"
• "Переведи это в VAST"

This skill is distinct from the analytical siblings. vast-validate checks a doc that already claims VAST shape and emits Pass/Warn/Fail. vast-okr-audit audits OKR-shaped docs for triad conflation. vast-transform takes prose that makes no VAST claim and produces the shape. If the input already has V/A/S/T headings and you want it scored, use vast-validate instead.

VAST recap (condensed)

The output sorts prose into four layers, each owning a distinct level of commitment:

Layer	Owns — what prose content maps here
Vision	The why / for-whom / enabled-experience. The purpose the work serves, as a falsifiable hypothesis. Look for "we believe", "so that", "for {users} who {pain}".
Architecture	The structural domains and the composition framework — skill library, interfaces, invariants, implementations. The durable structure work is built on. Look for "the system is structured as", "capabilities", "guarantees", "what must always hold".
Strategy	Investment direction and sequencing — which experiences to compose next, in what order, validated how. Look for "first we'll", "prioritize", "sequence", "invest in", "validate with".
Tactics	Concrete deliverables — specific compositions for specific users at specific moments. Look for named features, screens, dated artifacts, per-user delivery.

Two cross-cutting facts the transform respects: Vision is a committed falsifiable hypothesis (it needs named revision triggers — flag their absence as a gap, don't fabricate them), and the four layers form a commitment-priority order (Vision → Architecture → Strategy → Tactics), not a temporal sequence.

For the four layers, the triad (Outcomes / Use Cases / Outputs), and the invariants/implementations split, see ../../references/vast-essentials.md. For the four-scope model (company / product / function / WoW) and per-scope identifying signals, see ../../references/layer-definitions.md. The eight common confusions — which the produced output must not exhibit — are in ../../references/anti-patterns.md. The composition-depth spectrum (deep / medium / light / vocabulary-only) is in the framework's applicability.md. Detailed transform heuristics are in references/transform-heuristics.md — read those if any extraction or mapping is ambiguous.

Process

For the prose under transformation:

1. Identify scope — company / product / function / WoW. State it explicitly; it determines what each layer should contain and what "good" looks like. (Scope signals: references/transform-heuristics.md#scope-identification; the full per-scope signal lists are in ../../references/layer-definitions.md.)
2. Extract Vision — find the why / for-whom / enabled-experience in the prose. If it is genuinely present, lift and tighten it. If it is genuinely absent (the prose is all features/deliverables with no stated purpose), flag the absence — do not fabricate a confident Vision. Offer a candidate Vision clearly labeled ⚠️ CANDIDATE (confirm) that the user must confirm. (See references/transform-heuristics.md#vision-extraction.)
3. Map to Architecture / Strategy / Tactics — sort the remaining prose into layers: structural domains / composition framework → Architecture; investment direction / sequencing → Strategy; concrete deliverables → Tactics. Note thin or empty layers rather than padding them. (See references/transform-heuristics.md#layer-mapping.)
4. Recommend composition depth — deep / medium / light / vocabulary-only. Match depth to the work's actual AI-substrate engagement and scope; do not force deep where light fits (that is anti-pattern #6) nor claim vocabulary-only where deep is needed (anti-pattern #7). (See references/transform-heuristics.md#depth-recommendation.)
5. Surface gaps — what's missing for a complete VAST doc: no falsification triggers on the Vision, Architecture with no invariants, no for-whom in the Vision, no named accountable owner, scope ambiguity. (See references/transform-heuristics.md#gap-surfacing.)
6. Emit the proposed VAST doc per Output format below.

Heuristics

#	Heuristic	One-liner	Detail reference
1	Scope identification	Match prose against per-scope signals; pick ONE scope (company / product / function / WoW). If the prose mixes altitudes (company ambition + product-feature detail), pick the dominant scope and record the conflict as an open question — don't silently average.	references/transform-heuristics.md#scope-identification (primary); ../../references/layer-definitions.md (per-scope signals)
2	Vision extraction	Lift why/for-whom/enabled-experience if present. If absent, mark status ABSENT and offer a ⚠️ CANDIDATE (confirm) Vision derived from the deliverables — never present an inferred Vision as settled fact. Partial why → "partially inferred".	references/transform-heuristics.md#vision-extraction (primary); ../../references/anti-patterns.md (#1 Vision-as-Use-Case, #4 missing falsification triggers)
3	Layer mapping	Sort prose: structural domains / composition framework → Architecture; investment direction / sequencing → Strategy; concrete deliverables → Tactics. Don't pad a thin layer — note what's needed. Don't promote a feature to Architecture just to fill it.	references/transform-heuristics.md#layer-mapping (primary); ../../references/vast-essentials.md (triad; invariants/implementations)
4	Depth recommendation	Pick deep / medium / light / vocabulary-only by scope × AI-substrate engagement. One-line reason citing the signal that decided it. Don't force deep on light work (#6) or claim vocabulary-only where deep is needed (#7).	references/transform-heuristics.md#depth-recommendation (primary); ../../references/anti-patterns.md (#6, #7)
5	Gap surfacing	Enumerate what's missing for a complete VAST doc: Vision falsification triggers, Architecture invariants, for-whom, named accountable owner, scope clarity. Gaps are honest absences to report, not things to invent.	references/transform-heuristics.md#gap-surfacing (primary); ../../references/anti-patterns.md (#2 Architecture-by-default, #7 vocabulary-only)

Heuristic numbers above are stable identifiers, not an execution order. The Process runs scope → Vision → mapping → depth → gaps. Reference heuristics by name (e.g. "Vision extraction") if you annotate the output.

Output format

Produce a markdown document with this exact structure — a proposed VAST doc, not a score:

# Proposed VAST — {title} ({scope} scope)

## Vision
{extracted Vision — why / for-whom / enabled-experience. If genuinely absent,
write a single clearly-labeled candidate instead:
"⚠️ CANDIDATE (confirm): {inferred purpose} — derived from the deliverables below;
confirm or replace before use."}

## Architecture
{mapped structural content — domains, composition framework, invariants.
If thin or empty, state what the prose provides and what a complete Architecture
still needs (e.g. "no invariants stated — needs ≥1 quality + ≥1 fallback contract").}

## Strategy
{mapped investment direction / sequencing. If absent, note that no sequencing
logic was stated.}

## Tactics
{mapped concrete deliverables — features, screens, per-user delivery.}

---

## Notes
- **Vision status:** extracted | partially inferred | ABSENT (candidate offered — confirm before use)
- **Composition depth:** deep | medium | light | vocabulary-only — {one-line reason}
- **Gaps surfaced:** {bullets — what's missing for a complete VAST doc}
- **Open questions:** {bullets — what the author must decide}
- **Decisions deferred:** {bullets, if any}

Always include all five Notes fields. Where a field is empty, write (none) — keep every field present so the Notes block is identical in shape across clean and gap-heavy transforms (mirrors the exemplars' empty-state discipline).

The four layer sections always appear, even when a layer is thin or empty — an empty layer is itself a finding (state what's missing), not a section to omit.

Examples

See:

• ../../examples/07-transform-clean/ — prose with a clear latent Vision + identifiable A/S/T material → clean transform, Vision status "extracted".
• ../../examples/08-transform-no-vision/ — a pure feature/deliverable list with no stated why → A/S/T mapped, Vision status "ABSENT", labeled candidate offered (not fabricated).
• ../../examples/09-transform-mixed-altitude/ — prose mixing company-level ambition with product-feature detail → transform picks a scope, records the altitude conflict as an open question.