Generate a compact five-tier strategy portfolio when the user needs multiple materially different paths before execution. Implicitly invoke for explicit requests for options, alternatives, trade-offs, reframing, or help escaping a repeated stall/failure. Do not activate for direct implementation, single-answer advice, repository-evidence opportunity mining ($ideate), or detailed planning ($plan).
Installation
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Generate a compact five-tier strategy portfolio when the user needs multiple materially different paths before execution. Implicitly invoke for explicit requests for options, alternatives, trade-offs, reframing, or help escaping a repeated stall/failure. Do not activate for direct implementation, single-answer advice, repository-evidence opportunity mining ($ideate), or detailed planning ($plan).
Creative Problem Solver
Purpose: when a strategy choice is still open, generate a five-tier portfolio that compounds through an Artifact Spine, then stop for a human choice.
Activation boundary
Implicit invocation is enabled. The activation signal is an unresolved strategy choice plus a request for divergent options. Difficulty alone is not enough.
Activate when
The user explicitly asks for options, alternatives, trade-offs, a strategy portfolio, fresh angles, reframing, or "what else could we try?"
Progress is stalled or repeated attempts fail, and selecting a materially different approach is the next task.
A multi-constraint or high-uncertainty decision would benefit from several distinct conceptual frames before execution.
The user asks to brainstorm or ideate in the ordinary sense without requesting repository evidence mining, tickets, a detailed plan, or implementation.
The primary deliverable is at least three materially different paths and a human choice among them.
Do not activate; route elsewhere
Direct implementation, debugging, review, or execution after an approach is selected: use the normal task flow or the owning implementation/review skill.
A single factual answer, a straightforward recommendation, or comparison of a small known set where a portfolio would add ceremony: answer directly.
Evidence-backed repository or product opportunity mining, ranked improvement discovery, or choosing what to plan next: use $ideate.
Turning a selected direction into a detailed implementation plan, specification, or work graph: use $plan or $spec-pipeline as appropriate.
Architecture or codebase understanding without a request for divergent paths: use direct analysis or $codebase-archaeology.
The user already chose a tier or path and asks to execute it: hand off to the execution owner; do not regenerate the portfolio.
Tie-breakers
Explicit $creative-problem-solver invocation wins unless a safety or domain-specific boundary requires another owner.
$ideate wins when repository evidence mining and ranked opportunities are central to the request.
$plan wins when the direction is already selected and the requested output is an execution plan.
Activate this skill when the requested outcome is a choice set; do not activate when the requested outcome is one answer, one plan, or one implementation.
Contract (one assistant turn)
Name the current Double Diamond stage: Discover / Define / Develop / Deliver.
If Define is weak: propose a one-line working definition + success criteria, and treat the portfolio as learning moves.
Deliver options, then stop and ask for human input before executing.
Always include a five-tier portfolio: Quick Win, Strategic Play, Advantage Play, Transformative Move, Moonshot.
Before generating tiers, name the default solution basin and ensure at least two tiers come from different conceptual frames.
For each option: accretive artifact + expected signal + escape hatch.
Run an Aha Check after reframing.
Keep a short Knowledge Snapshot (facts/risks/assets) + Decision Log.
Define (converge): lock the problem statement + success criteria; surface unknowns and ask.
Develop (diverge): generate solution paths; prototype/experiment if needed.
Deliver (converge): pick a tier to execute; hand off to tk for incision + proof.
Mode check
Pragmatic (default): ship-this-week options only.
Visionary: only when asked for long-horizon strategy or systemic change.
Lane selector
Fast Spark: skip broad candidate generation; produce the portfolio directly.
Full Session: diverge (10-30 ideas), cluster, score, then select one option per tier.
Reframe selection (required)
Supported techniques:
Discover default -> Assumption Mapping
Define default -> How Might We
Develop default -> SCAMPER
Deliver default -> Pre-mortem
Final fallback -> First Principles
Rule: start with the stage default. If there is no Aha, run exactly one second pass with First Principles.
If the user asks for an out-of-catalog technique, choose the nearest supported technique and disclose the supported Reframe used.
Chat disclosure: include only Reframe used: <technique> + a one-line why.
Aha Check (required)
Definition: a restructuring insight (new representation/model).
Output: one-line insight. If none after the second pass, state N/A after second pass and continue with the most conservative portfolio that still satisfies the contract.
Portfolio rule
Every response must include all five tiers.
If stage is Discover/Define (problem unclear), the tiers are learning moves (not build proposals).
If stage is Develop/Deliver (problem clear), the tiers are solution moves.
Diversity guard
Name the default solution basin before generating tiers.
Ensure at least two tiers come from different conceptual frames, not just different sizes of the same idea.
Valid frame shifts include substrate, interface, constraint, proof surface, incentive, time horizon, operating model, or user ritual.
If the five tiers cannot honestly diverge, say so and mark the portfolio as same-basin.
Accretion (required)
Accretive artifact: a durable asset you keep even if the option is wrong (measurement, harness, spec, test, automation, interface, dataset, doc).
Rule: every tier must name one accretive artifact.
Artifact Spine: define 1-3 named artifacts shared across tiers; each tier's artifact must be one of these or an extension of one (otherwise explicitly state why it is a different spine).
Spine output: for each spine artifact, include purpose + minimal shape/interface + timebox + where it lives (repo path or conceptual home).
Ladder: prefer stacking (Quick Win builds the base; higher tiers reuse/extend it). If it does not ladder, say so explicitly.
