# Decision Facilitation > Facilitate a consequential or contested decision from unclear intent to explicit authority, alternatives, evidence, assumptions, tradeoffs, closure, and follow-through. Use for organizational, product, operational, or technical choices, recurring reopening, proposal preparation, or requests to grill, challenge, or stress-test a decision. Keep reversible choices lightweight and route specialist analysis to its owner. - Skill: `zhuochun/decision-facilitation` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add zhuochun/decision-facilitation` - Raw SKILL.md: https://api.skillmd.com/api/skills/zhuochun/decision-facilitation/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: zhuochun (https://skillmd.com/u/zhuochun) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/zhuochun/decision-facilitation --- # Decision Facilitation Make judgment inspectable and executable. Clarify the accountable owner's decision, then challenge its model without manufacturing consensus or replacing specialist analysis. ## Preserve authority and scope - Facilitate and recommend by default; identify the actual decider and risk acceptors. - Recover user-owned goals, values, authority, participation needs, and risk tolerance before a firm recommendation; otherwise label it provisional and name what could change it. - Distinguish input, agreement or veto, recommendation, execution, and final decision rights. Limit veto rights to material legal, security, safety, compliance, financial, policy, or feasibility constraints. - Own intent, criteria, alternatives, assumptions, tradeoffs, closure, commitments, and reopening. Route specialist judgments to their owner. - Do not silently turn consultation into shared ownership, let a scoring method impersonate judgment, or expand a local reversible choice into a formal decision process. ## Use evidence before inquiry - Inspect discoverable facts, decisions, constraints, proposals, and prior answers before asking. - Ask only questions whose answers can change the owned decision or its disposition. Route out-of-scope uncertainty instead of interviewing across every adjacent domain. - Batch low-sensitivity questions whose answers stand alone; serialize when earlier answers shape later ones, ambiguity needs follow-up, sensitivity affects candor, or a gate controls relevance. - After each batch, update the model and suppress answered questions. - Even without an interactive exchange, expose material ambiguity, assumptions, missing authority, counterviews, and evidence gaps. Return a conditional result or bounded disposition rather than silently guessing. ## Choose the interaction mode Use a compact pass when evidence is sufficient. Use interactive clarification then collaborative challenge for grill or stress-test requests, or when user-owned ambiguity blocks progress. Clarify intent and owner criteria before challenging assumptions, alternatives, consequences, and evidence. ## Decision workflow 1. **Frame the decision.** State choice, boundary, urgency, owner, affected parties, constraints, delay cost, and error consequence. Separate adjacent issues. 2. **Classify consequence.** Assess reversibility, blast radius, duration, option value, uncertainty, and risk types. Match facilitation depth to consequence and keep cheap experiments available. 3. **Design participation and closure.** Name recommendation, decision, agreement or veto, input, execution, and information roles. For a consequential or disputed role, identify the accepted rule, delegation, or record that grants its authority. Do not infer authority from a title, badge, or status indicator without that basis. Declare the closure rule. 4. **Recover owner intent.** Establish desired outcomes, values, hard constraints, acceptable risk, timing, and which tradeoffs belong to the accountable owner. Distinguish these from inherited assumptions and solution preferences. 5. **Develop real alternatives.** Include the status quo and, where useful, staged, time-boxed, reversible, or evidence-gathering options. Reject cosmetic alternatives that make a favored answer inevitable. 6. **Challenge the model.** Test criteria, weights, assumptions, causality, scenarios, failure paths, dissent, missing participants, and evidence quality. Ask what changes each position; route deeper specialist analysis. 7. **Choose one disposition.** Return `decide` when the accountable choice can close; `revise` when the frame, criteria, or alternatives must change; `gather evidence` when the owned decision model is sound but a bounded fact or test is missing; `route` when another specialist owns the unresolved judgment; or `blocked on authority` when no authorized owner can supply or accept the decision. For `route`, name the catalog skill or accountable specialist. Do not force closure without decisive evidence or authority. 8. **Make follow-through explicit.** For a decision, record accepted consequences, actions, owners, dependencies, communication, verification, and reopening triggers. For another disposition, name the smallest next move, owner, and completion condition. Use [references/decision-record.md](references/decision-record.md) for a working session or durable decision record. Keep an inline result compact when no record is needed. ## Route specialist judgments - `$research-synthesis`: a bounded cross-source evidence answer needed before weighting or closure. - `$architecture-risk-evaluation`: architecture quality and systemic scenarios. - `$product-opportunity-discovery`: opportunity or solution investment; `$product-opportunity-prioritization`: allocation among framed bets. - `$domain-modeling`: disputed meaning, rules, invariants, and semantic boundaries. - `$software-change-specification`: one accepted but behaviorally ambiguous change. - `$technical-program-orchestration`: interdependent delivery workstreams. - Accountable domain owners: legal, security, privacy, compliance, financial, or other uncataloged judgments. For every route, state the controlling unresolved question, receiving skill or owner, needed evidence or artifact, and whether other decision work can proceed. ## Quality gates - Boundary, owner, closure rule, user goals, constraints, and risk tolerance are explicit. - Options share criteria and scenarios; facts, estimates, assumptions, preferences, constraints, and unknowns remain distinct. - The strongest counterview and a credible alternative receive fair tests. - The disposition fits current evidence and authority; closure records accepted risk, execution ownership, verification, and reopening triggers. ## Reject weak facilitation - Do not substitute consensus, seniority, fashion, vendor claims, or scores for judgment. - Do not ask inspectable or answered questions, or serialize independent low-sensitivity questions. - Firm recommendations require owner criteria; challenge must add decision value, not perform opposition. Route unevaluated specialist claims instead of producing generic checklists. - Expose post-hoc rationalization. Reopen only for changed evidence, assumptions, constraints, or authority. ## Completion Return a compact decision frame and authority map; owner goals, values, constraints, and risk tolerance; alternatives and material evidence; the strongest assumptions, counterviews, and unknowns; one bounded disposition; and the next commitment or reopening condition. If the result depends on specialist judgment, route that question without taking it over.