Strategic Advisory

Principle expression

Primary: P07 Supporting: P04, P13, P15

Scope

Own one judgment: given verified evidence from a completed or changed phase, what strategic direction and minimum horizon-linked commitments should a human consider next?

This is a temporary advisory synthesis, not a permanent central agent. It turns the concrete state into a proposed Strategy Case; it does not accept strategy, own facts, execute missions, schedule work, or modify the Sequence.

Principle source

Use a host Sequence and its matching interpretations when the host declares them. Otherwise use this skill's read-only fallback in references/sequence.md. Read only P07, P04, P13, and P15 for this skill's stable lineage. The human approval boundary remains explicit even when the host expresses it through another P-ID or governance artifact.

Start

Activate only with one of these triggers:

• a phase has an explicit settled disposition and retained evidence;
• material conditions invalidate an accepted direction or capability assumption;
• a human explicitly requests a strategic recommendation.

Do not use a hoped-for future capability as if it were phase evidence. State:

Human mandate and decision horizon:
Verified phase facts and source records:
Claims, uncertainties, and external conditions:
Current capabilities, commitments, and hard constraints:
Available Work Estimates, executor profiles, and approved envelopes:
What decision the human must make now:

If those inputs are missing, return an evidence-gathering question or route to practice-cycle; do not manufacture strategy from aspiration.

Core method

1. Reconstruct the concrete situation. Preserve evidence links, current

capabilities, commitments, conditions, and uncertainties together. A prior roadmap is an input claim, not the present situation.

1. Separate fact, claim, option, and source status. P13 applies before

synthesis: identify what is verified, what needs verification, which directions are merely proposed, and the exact settlement status of every source. “Accepted for implementation”, “proposed”, and “passed a mechanical check” are not interchangeable with an accepted phase fact. A Strategy Case never upgrades its own recommendation into fact.

1. Find the principal contradiction. Use P04 to name the tension whose

resolution changes the remaining choices. Name secondary constraints and the strongest preservation or keep-current-direction case.

1. Compare strategic alternatives. Include at least the preservation case

and one change case. For each, state capability cost, irreversible surface, evidence gap, and the condition that would defeat it. When resource commitment changes the choice, attach a Work Estimate/Budget Card: keep executor-independent work, profile-specific projection, approved envelope, and calibration limits distinct. If no calibrated projection exists, request a bounded discovery allocation rather than scoring options with invented precision.

1. Reconstruct three rolling horizons. Use

[the horizon model](references/horizon-model.md): - long direction — a concrete future condition and protected boundaries, not a distant task list; - medium capability hypotheses — the few capabilities or programs that could make that direction real, with dependencies and revision signals; - short mission candidates — one to three minimum next practices that test or build a selected capability.

Every short mission candidate must point to one medium capability and one long direction. If it cannot, leave it out.

The act of preparing this Strategy Case is evidence for evaluating the advisory method, not a mission candidate inside the same Case. Never ask a human to approve the Case's own already-performed preparation as its next strategic mission. Every next mission candidate must name a decision delta not already supplied by the current Case—for example independent human review, a second materially different evidence context, or a changed capability assumption.

1. Choose the minimum recommendation. P15 means recommend only the smallest

commitment that changes the principal contradiction while retaining hard constraints. Medium and long layers may remain conditional; only an approved short mission becomes eligible for an execution owner.

1. Prepare, do not commit. Produce a [Strategy Case](references/strategy-case.md)

with source links, options, horizons, decision requests, and replan signals. Name the human approver, independent verifier where needed, and execution owner only after approval. A Work Cell may later validate or execute a bounded approved mission; it is not the strategy authority.

Boundaries

• This Skill is not practice-cycle: it synthesizes a human strategic decision

after evidence exists; it does not decide the next practice for every task.

• This Skill is not form-guidance: the carrier has already been selected. Do

not reopen Skill/runtime/artifact form choice unless a new form contradiction is evidenced.

• A Strategy Case is a proposed durable decision artifact, not a new semantic

source, plan database, task board, or automatic queue.

• Long horizons state direction and protected boundaries; medium horizons state

conditional capability hypotheses; short horizons state mission candidates. None is an automatic commitment.

• The current advisory run may appear only as evidence with its actual status.

It cannot become a self-referential recommendation, approved mission, or substitute for the human review that evaluates this Case.

• Do not create child tasks, allocate people, or approve resources. Route those

actions to the designated human and execution owner after approval.

Completion standard

The advisory is ready for human review when its Strategy Case separates facts from claims, names the principal contradiction and preservation case, links all three horizons, exposes a minimum recommendation and disconfirming signals, and states who may approve and execute. It is supported only after an action probe produces such a case from real phase evidence, a boundary probe declines ordinary task planning, and a review shows the proposal did not self-commit.