| name | decision-package |
| description | Produce the Decision Package — the paid keystone deliverable of a semantic/data consulting engagement's P1 Decision Phase. Use to write the recommendation, the in/out scope, the sequenced (pilot → scale) roadmap, the buy/build/defer decisions, and the ready-to-contract execution brief that hands off to the build/architecture tier. The unit of value is decision-readiness; the artefact sits ABOVE architecture (it justifies and scopes the build before a repo exists). Triggers — "produce a decision package", "write the recommendation / scope / roadmap / buy-build-defer", "decision-readiness deliverable", "execution brief for the build", "scope this engagement", "Semantic Readiness & Direction / Decision Foundation deliverable". This skill PRODUCES the artefact; it does NOT coach the engagement design or hold the free→paid line (that is semantic-consulting-coach — "produce" vs "coach"). Not for free P0 orientation, not for delivery artefacts (an ontology, a mapping), not for code architecture. |
| license | Apache 2.0 |
| metadata | {"category":"consulting"} |
Decision Package
Provisional pending the first-engagement gate. The consulting depth here refines after the first real
paid P1 engagement; treat this wiring as the floor, not the finished standard.
Overview
The Decision Package is the paid keystone deliverable of the consulting engagement's P1
Decision Phase. Its unit of value is decision-readiness: the state in which the client knows
what to do next, why, what success looks like, and what comes later — the moment uncertainty
becomes safe to commit. This skill PRODUCES that artefact.
It sits deliberately above architecture: it justifies and scopes the build before a repo
exists. Architecture, the conceptual model, and the EPICs all descend from it — the Decision
Package is the consulting-tier root of the golden thread.
Provenance. The coach (semantic-consulting-coach) shapes whether and how to run P1 and holds
the free → paid boundary; the design intent of the Decision Phase lives in its
engagement-model.md. This skill picks
up after that line is crossed and the client has bought P1: it builds the deliverable. Keep the
boundary crisp — produce here, coach there.
The deliverable: five parts (R1)
The Decision Package is an executive artefact framed around decision-readiness. Frame and write
it with the executive-communication skill
(Governing Thought → SCQA → Minto pyramid, answer-first); this skill supplies the what, that skill
supplies the how it reads. Its five required parts:
- Recommendation — the first initiative to commit to, stated answer-first as a Governing
Thought, with the reasoning that makes it safe to commit.
- Scope — explicitly in and explicitly out. The "out" list is load-bearing: it protects
the boundary with the build tier (P2) and prevents scope creep.
- Sequenced roadmap — pilot → scale: what to do first, then next, in deliberate order.
- Buy / build / defer decisions — for each capability, a named decision with rationale.
- Ready-to-contract execution brief — the hand-off to the build tier / architecture: enough for
P2 (or another implementer) to scope and contract cleanly.
It earns the right to say, calmly: "From here, we can execute, or you can take this and execute with
someone else." The full fillable template is in
references/decision-package-template.md.
The P1 discovery flow (R3, R5)
The package is the output of a structured flow, not a maturity assessment or open-ended discovery.
This is one coherent discovery framework — the still-relevant items of the internal
product-blueprint checklist are absorbed into it where they serve decision-readiness, and the
pure product-delivery mechanics are dropped (the reconciliation is in
references/blueprint-absorbed.md).
- Structured discovery — the problem, the deciding persona, the measurable outcome, the
competitive/strategic "why now". (Absorbs blueprint Q1 problem/persona/outcome, Q2 success
metrics, Q3 why-we-win — reframed as decision inputs, not an MVP brief.)
- Landscape / data reading — the data estate, governance maturity, legacy constraints, existing
assets and obligations.
- Gap analysis — current state vs. the strategic ambition: where the gaps are, which ones
block the first safe step, which can wait.
- Option framing — the candidate first initiatives, each with its trade-off (the "big
decision" framing of blueprint Q4, kept as options to decide between, not a pre-made choice).
- Sequencing — order the options into a pilot → scale roadmap; name what is explicitly not
in the first step (the discipline of blueprint Q6/Q7, recast as roadmap + out-of-scope).
- Buy / build / defer — for each needed capability, decide and justify.
- Execution brief — assemble the ready-to-contract hand-off to the build tier.
Detail, including gap-analysis and buy/build/defer guidance, is in
references/discovery-flow.md.
First-cut conceptual model (R4)
A first-cut conceptual model of the domain often surfaces during discovery (a sketch of the key
entities and relationships, to ground scope and options). When it does, produce that fragment with
the conceptual-modelling skill — but keep it a
sketch in service of the decision. Do not pull modelling depth (LinkML authoring, generation,
ontology engineering) into the Decision Package; the real model is built later, in the build tier.
The free → paid boundary (R6)
| Phase | Question | Commercial | Owner |
|---|
| P0 Orientation | "Is this even relevant to us?" | Free, shallow | semantic-consulting-coach |
| P1 Decision | "What do we do first, and why?" | Paid | this skill |
This skill operates only inside paid P1. Orientation is free, shallow, and never deeply
customised — that is the coach's territory, and the coach protects the line. Do not blur Decision
Package production into free coaching: if the work being asked for is still orientation, it has not
crossed the boundary, and this skill should not be producing a paid artefact yet.
Golden-thread root: where the package lives (R7)
The durable Decision Package lands as its own spine artefact type, at:
openspec/decisions/<id>.md
It is not forced into openspec/specs/ — those are RFC-2119 behaviour contracts, and a Decision
Package is an executive narrative, a different genre. It gets its own home so the genres do not
collide.
It is the consulting-tier root of the golden thread (see
../../spine/golden-thread.md). The first specs/ /
architecture entries cite it as their parent:
decision-package (openspec/decisions/<id>.md)
→ requirement / architecture
→ conceptual model
→ EPIC / change
→ task → test → commit
Make this convention explicit when producing the package: mint a human-readable ID, and ensure the
downstream architecture/requirement artefacts cite it as parent.
Naming (R8)
The default name is "Decision Package" — for both this skill and the client-facing deliverable.
An engagement MAY relabel the client-facing artefact without renaming the skill, e.g.
Semantic Readiness & Direction, Decision Foundation, Scope Definition. This is an allowed
per-engagement override: the skill name and the five-part structure stay fixed; the cover label
flexes to the engagement.
Boundary & Related Skills
This skill OWNS: producing the Decision Package (the five R1 parts) and running the P1
discovery flow that yields it (structured discovery → landscape reading → gap analysis → option
framing → sequencing → buy/build/defer → execution brief), including the absorbed product-blueprint
items recast for decision-readiness.
This skill DELEGATES:
Related: semantic-consulting-coach, executive-communication, conceptual-modelling,
architecture.
Tone & style
British English. The Decision Package is an executive artefact: answer-first, concise, every claim
load-bearing. Recommendation before proof; scope boundaries and trade-offs explicit; buy/build/defer
decisions named, not hedged. Write it in the engagement's executive voice (see
executive-communication), not consultancy-bland.