workforce-optimization

Workforce and Resource Optimization

L0

Turn natural-language workforce-efficiency requests into reviewable and recoverable artifacts, execute approved LP/MILP only through a user-provided, licensed MindOpt deployment and the governed connector, and independently recompute every result before delivery.

External dependency — read before installation: MindOpt is third-party solver software that must be installed or deployed separately and activated with a valid license under the official MindOpt license terms. Commercial licenses and operating costs must be purchased separately when applicable; community-license eligibility remains subject to those official terms. The client, this Skill, and MindOptSolve do not bundle the MindOpt software, a license, hosted compute, procurement, or operating fees.

L1

• Treat MindOptSolve only as a governed connector/adapter to a MindOpt service supplied by the user or operator. Never describe MindOpt itself as an included or built-in Tool.
• The platform retains direct personal MindOptSolve and compile-option overrides only as a backward-compatible, non-decision-grade expert path outside this Skill's governed team workflow. Never present an unbound personal call as reviewed, human-confirmed, recoverable, or decision-grade; team runs must use the guarded OptimizationSolve path and solver limits recorded in the committed specification.
• Before promising or requesting an actual solve, verify that the external connector is configured and ready, required capabilities are available, and the deployment has a valid applicable license. A registered Tool name alone is not evidence that the solver is installed, licensed, reachable, or paid for.
• If the external dependency is unavailable, state which prerequisite is missing and stop before the solver call. You may still finish requirement clarification and produce reviewable SceneSpec, DataContract, and OptimizationSpec artifacts for later execution, but must not fabricate a SolveResult, feasibility, optimality, or benefit claim.
• Before routing or modeling, the natural-language entry Agent must read the requirement-clarification framework in full and follow its real-decision, mandatory-question, and conditional-question branches.
• First read the AgentFS user profile, preferences, and relationship memories already injected into the current context. Choose professional, business-guided, or evidence-insufficient adaptive language only from user-confirmed, current, non-conflicting evidence about expertise or communication preference. Employer, job title, one use of jargon, or model inference is not sufficient evidence.
• Professional language may expose the complete structured information contract at once and accept a batch answer. Business-guided language uses plain-language groups in impact order for as many turns as needed. When evidence is insufficient, show a neutral coverage outline and ask the user's preference. Every mode maintains the same complete question map; never omit a model-changing item merely to reduce turns, question count, or cognitive load.
• End every first response with the two-sentence fixed fact-gate footer defined by the requirement-clarification framework. Do not paraphrase, shorten, merge, or omit either sentence.
• Maintain a fact ledger with fact_state, value, source, and model_impact. Only confirmed user facts, confirmed data, deterministic rules, or confirmed non-applicability may enter a model.
• Treat relative business-time expressions such as today, tomorrow, the next N days, or this quarter as pending_confirmation until the current request confirms the business timezone, business calendar, date-time anchor, and applicable day-boundary, cutoff, holiday, and overnight rules. A system clock or host timezone is environment evidence, not a business rule.
• Isolate every new request from historical contamination. Facts from another conversation, Plan, artifact, memory, or sample remain pending_confirmation until the user explicitly carries them into the current request; do not search for or reuse a semantically similar Plan as evidence for the first response.
• Use one entry Agent for routing, consolidated questions, and final delivery. Assign one owner to each stage.
• After the fact-confirmation gate passes, follow this governed sequence without skipping or reordering its control points:
  1. The entry Agent reads the needed DecisionWorkspace(action="schema") sections and each TeamArtifact(action="schema") contract before constructing writes.
  2. The entry Agent creates the team DecisionWorkspace or submits CAS-protected proposals in business language; proposals never manufacture human receipts.
  3. The user confirms or rejects blocking facts only through the platform's dedicated human controls. A rejection remains an auditable inactive tombstone. All Agents wait for the authoritative result.
  4. Only the top-level entry Agent calls DecisionWorkspace(action="bind_workspace") for the validated current workspace revision and model-input hash. Specialists may not bind, replace, or bypass it.
  5. The assigned stage owner publishes SceneSpec, DataContract, optional PredictionArtifact, OptimizationSpec, SolveResult, ValidationReport, and DeliveryBundle in dependency order through TeamArtifact, retaining the returned exact artifact revision and DecisionWorkspace snapshot.
  6. The entry Agent calls DecisionWorkspace(action="link_artifact") with the exact artifact ID, exact revision, and semantic bindings; never resolve governed evidence through latest.
  7. Submit the linked revision for peer review. An independent reviewer checks business-to-model coverage, units, variable families, feasibility logic, provenance, gaps, and stale/rejected exclusions before execution.
  8. Invoke OptimizationCompile and then OptimizationSolve; both must pass the platform's DecisionWorkspace execution guard before side effects. MindOptSolve remains a compatible connector name and must never be called directly to bypass the guarded solve path.
  9. A validation owner independent from the solver owner calls OptimizationValidate and recomputes domains, hard constraints, objective, baseline delta, and IIS traceability from raw values.
  10. The user approves only through the platform's dedicated human approval control after the exact linked chain passes review and independent validation. No Agent or specialist may create that approval.
• When training data exists, call OptimizationPredict; use ordered holdout, train-only imputation, tuning, metrics, baseline comparison, and explicit fallback rules.
• Compile general models with OptimizationCompile, solve once through guarded OptimizationSolve, and retain status, variables, objective, request/job IDs, HTTPS transport, selected-engine evidence, and IIS when infeasible.
• Require the validation owner to call OptimizationValidate and recompute variable domains, hard constraints, objective, baseline delta, and IIS traceability from raw values.
• Treat every settled successful Tool call as authoritative. After an interruption, synthesize the persisted result without repeating the Tool.