# Scenario Planning > Use when the user wants scenario planning for an uncertain future: framing a focal question, finding critical uncertainties, building a 2x2 of plausible futures, and deriving strategy that holds across them. Gathers driving forces and selects the two critical uncertainties as axes, then fans out four scenario-builder subagents in parallel (one per quadrant) to develop named narratives, and synthe… - **Type:** Skill - **Install:** `agentstack add skill-l4ci-latticework-scenario-planning` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [l4ci](https://agentstack.voostack.com/s/l4ci) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [l4ci](https://github.com/l4ci) - **Source:** https://github.com/l4ci/latticework/tree/main/skills/scenario-planning ## Install ```sh agentstack add skill-l4ci-latticework-scenario-planning ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # scenario-planning Builds several distinct, plausible futures for a focal question using the 2x2 critical-uncertainties method (Global Business Network / Royal Dutch Shell tradition; Peter Schwartz, *The Art of the Long View*, 1991). The goal is to prepare for an uncertain future, not to predict which one arrives. The method, the axis-selection logic, what makes a good axis and a good scenario, how to write signposts, and the pitfalls live in [references/scenarios.md](references/scenarios.md). Load that file before starting and follow it exactly. ## When to use The user faces a decision whose right answer depends on an uncertain future and wants to stress-test strategy against several of them rather than a single forecast. The unit of analysis is a focal question tied to a decision-maker and a horizon, not "the future" in the abstract. Scenario planning complements [pestel-analysis](../pestel-analysis/README.md), which is the natural source of driving forces (its high-impact, high-uncertainty watch list feeds Stage 1 directly), and [three-horizons](../three-horizons/README.md), which sequences moves over time once the scenarios are set. If the user has already run a PESTEL scan, take its forces as input rather than regathering them. ## Language Conduct the session and write the report in the language the user is writing in. The reference file is English; keep the framework and its terms keyed to it regardless of the output language. ## Inputs - **FOCAL QUESTION** (required): the decision or strategic question, and the entity it concerns. Push for specificity: "what will our market look like" is too loose until tied to a decision-maker and a choice. Narrow it with a question or two if vague. - **HORIZON** (optional): how far out to look. Default to a sensible multi-year horizon (five to fifteen years), far enough that today's certainties could break but not so far it turns to fiction. - **CONTEXT** (optional): driving forces already known (for example from a prior PESTEL scan), and the decision the scenarios should inform. ## Orchestration map ``` Stage 1 Forces + Axes ── gather driving forces, then SELECT 2 critical uncertainties ──┐ Stage 2 Scenarios ── 4 scenario-builder subagents IN PARALLEL, one per quadrant ───┤ Stage 3 Synthesis ── 1 agent, needs all 4 ─────────────────────────────────────────┘ (robust vs contingent, signposts) ``` Tell each subagent its final message is the return value: structured data, not prose for a human. Subagents should use available retrieval to ground forces and scenario logic in real, current developments. ## Stage 1: Force gathering and axis selection Gather the driving forces first, then select the two axes. Both steps happen before any scenario is built. **Gather the forces.** If a PESTEL scan or a force list is supplied in CONTEXT, use it. Otherwise cast wide across political, economic, social, technological, environmental, legal, and the relevant industry forces. You may fan out a few force-finder subagents by lens to gather concurrently, or do it in one pass; either way produce a single deduplicated list of driving forces that could shape the focal question. **Rank and select (this stays with the orchestrator).** Score every force on two dimensions against the focal question, per the impact-x-uncertainty matrix in the reference file: - **Impact**: how much the force would move the answer to the focal question. - **Uncertainty**: how unpredictable its trajectory is over the horizon. Sort the forces into the four cells. Set aside the **predetermined elements** (high impact, low uncertainty) as the shared backdrop every scenario inherits. From the **critical uncertainties** (high impact, high uncertainty) pick the two most decisive that are also independent of each other, and render each as an axis: a spectrum between two opposed, plausible poles. Run the independence and goodness tests in the reference file before committing; if the two axes are correlated, the 2x2 collapses into one future at four intensities. Crossing the two axes defines the four quadrants. Name each quadrant's pole combination so Stage 2 knows exactly which world to build. Confirm the two axes with the user before fanning out, since everything downstream rests on this choice. ## Stage 2: Scenario building (PARALLEL, four scenarios) Dispatch four scenario-builder subagents at once, one per quadrant of the 2x2. Give each the shared framing plus its quadrant's pole combination and the predetermined elements that every scenario shares. **Shared framing (send to every builder):** > Build one scenario for this focal question: **[FOCAL QUESTION]**, over **[HORIZON]**. Your scenario sits at this corner of the 2x2: **[AXIS A pole] x [AXIS B pole]**. Every scenario shares this backdrop of predetermined elements: **[predetermined elements]**. Develop a coherent, internally consistent world, not a bullet list. Return: > - **Name**: a vivid, memorable name for this world > - **The path here**: how the world got from today to this corner over the horizon, with the key turning points > - **What it looks like**: the state of the world in this scenario, with the predetermined elements playing out inside it, grounded in real evidence where retrieval allows > - **Implications for the focal question**: what this world means for the entity and its decision > - **Distinctive signals**: the early, observable indicators that this world specifically is emerging (not the other three) > - **Confidence**: high, medium, or low on the internal logic, lowered where evidence was thin > > Treat this as a plausible future to prepare for, not a prediction. Do not assign it a probability. Context: [CONTEXT] Assign the four quadrants by their pole combinations. Collect all four structured scenarios before continuing, and re-dispatch any builder that fails rather than synthesizing with a gap. ## Stage 3: Synthesis (SEQUENTIAL, needs all 4) One agent turns the four scenarios into strategy: 1. **The scenario set at a glance.** The 2x2 with the four named scenarios in their quadrants, and a line on each. Check the set: are the four genuinely distinct worlds, or the same world at four intensities? If they collapse, the axes were wrong; flag it. 2. **Robust (resilient) moves.** The actions that make sense across all four scenarios. These are the core of the strategy because they carry no bet on which future arrives. 3. **Contingent moves.** Hedges and options tied to specific scenarios, each held ready rather than committed now, and each wired to the signpost that would trigger it. 4. **Signposts to monitor.** The early-warning dashboard: observable, discriminating, leading indicators that show which scenario is emerging. Pull the distinctive signals from each scenario and sharpen them into a watch list, noting which scenario each one points toward. 5. **Implications for the decision.** What the focal question's decision-maker should do now (the robust moves), what to prepare (the contingent moves), and what to watch (the signposts) as the future declares itself. ## Report structure Write a thorough markdown report and save it to `scenario-planning--.md` (today's date) in the working directory unless the user names another location. 1. **Header.** The focal question, the entity, the horizon, the date, and a one-line note on the method and that scenarios are not forecasts. 2. **The 2x2 at a glance.** The two axes, the four named scenarios placed in their quadrants (a simple ascii 2x2 is fine), and one line on each. A reader with two minutes should leave knowing the four futures and what to do across them. 3. **Axes and forces.** The two critical uncertainties chosen as axes and why; the predetermined elements forming the shared backdrop; and the forces considered but set aside. 4. **Per-scenario sections (four).** For each: name, the path from today, what the world looks like, and the implications for the focal question. Grounded in specifics, not generic futures. 5. **Strategy.** Robust moves up front, then contingent moves each tied to its scenario and trigger. 6. **Signposts.** The monitoring dashboard: indicator, which scenario it points toward, and why it discriminates. 7. **Caveats.** Scenarios are plausible futures to prepare for, not predictions, and carry no probabilities; the set is only as good as the two axes; the value is in the strategy and signposts, not the stories; and the scenarios are a tool for thinking, to be revisited as forces shift. ## Principles - **Prepare, do not predict.** The output is strategy that holds across futures, not a guess at which one arrives. No scenario gets a probability. - **Axes must be uncertain and independent.** Only high-impact, high-uncertainty forces earn an axis, and the two must be uncorrelated. Predetermined trends are the backdrop, never the cross. - **Four distinct worlds, not four intensities.** Each quadrant must demand a different response. If the scenarios are one future dialed up and down, the axes were wrong. - **End in strategy.** Robust moves, contingent hedges, and the signposts that trigger them. Vivid narratives with no strategy are entertainment. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [l4ci](https://github.com/l4ci) - **Source:** [l4ci/latticework](https://github.com/l4ci/latticework) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-l4ci-latticework-scenario-planning - Seller: https://agentstack.voostack.com/s/l4ci - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.