金大哥 - world-model-workflow Skill 详情

Intent

Run the composed workflow world-model-workflow using atomic capability skills to construct a comprehensive, grounded representation of a system or domain.

A world model captures:

State: Current entity states and attributes
Dynamics: How the system evolves over time
Uncertainty: Confidence bounds and unknowns
Provenance: Source and lineage of all facts

Success criteria:

Complete entity inventory with identity resolution
State representation follows canonical schema
Causal relationships and dynamics modeled
Uncertainty quantified for all assertions
Full provenance chain for every fact
Simulation capability established

Compatible schemas:

reference/world_state_schema.yaml
reference/event_schema.yaml

Inputs

| Parameter | Required | Type | Description | |-----------|----------|------|-------------| | goal | Yes | string | The modeling objective (e.g., "model supply chain for disruption analysis") | | scope | Yes | string|array | Domain, system, or entities to model | | constraints | No | object | Limits (e.g., time horizon, resolution, confidence threshold) | | sources | No | array | Data sources for world state extraction | | prior_model | No | object | Existing model to extend or refine |

Procedure

Create checkpoint marker if mutation might occur:
- Create .claude/checkpoint.ok after confirming rollback strategy
Invoke /retrieve and store output as retrieve_out
- Gather raw data from configured sources
Invoke /inspect and store output as inspect_out
- Examine retrieved data for structure and quality
Invoke /identity-resolution and store output as identity-resolution_out
- Resolve entity references and establish canonical IDs
Invoke /world-state and store output as world-state_out
- Construct canonical state representation
Invoke /state-transition and store output as state-transition_out
- Define rules for state evolution
Invoke /causal-model and store output as causal-model_out
- Map cause-effect relationships
Invoke /uncertainty-model and store output as uncertainty-model_out
- Quantify confidence and unknowns
Invoke /provenance and store output as provenance_out
- Document source and lineage of all facts
Invoke /grounding and store output as grounding_out
- Attach evidence anchors to assertions
Invoke /simulation and store output as simulation_out
- Validate model through simulation runs
Invoke /summarize and store output as summarize_out
- Generate human-readable model summary

Output Contract

Return a structured object:

workflow_id: string  # Unique model construction ID
goal: string  # Modeling objective
status: completed | partial | failed
world_model:
  version: string
  created_at: string  # ISO timestamp
  schema_version: string
  entities:
    count: integer
    by_type: object  # type -> count
    sample: array[object]  # representative entities
  relationships:
    count: integer
    types: array[string]
    sample: array[object]
  evidence_anchors: array[string]
state:
  snapshot: object  # Canonical world state
  hash: string  # Integrity hash
  timestamp: string
  evidence_anchors: array[string]
dynamics:
  transition_rules: integer
  causal_links: integer
  temporal_scope: string  # e.g., "real-time", "daily", "event-driven"
  evidence_anchors: array[string]
uncertainty:
  overall_confidence: number  # 0.0-1.0
  high_uncertainty_areas: array[string]
  unknown_factors: array[string]
  evidence_anchors: array[string]
provenance:
  sources: array[string]
  lineage_depth: integer
  coverage: number  # 0.0-1.0 (% of facts with provenance)
  evidence_anchors: array[string]
simulation:
  validated: boolean
  scenarios_tested: integer
  anomalies_found: array[string]
  evidence_anchors: array[string]
summary:
  description: string
  key_insights: array[string]
  recommended_actions: array[string]
  evidence_anchors: array[string]
confidence: number  # 0.0-1.0
evidence_anchors: array[string]
assumptions: array[string]

Field Definitions

| Field | Type | Description | |-------|------|-------------| | workflow_id | string | Unique identifier for this model construction | | world_model | object | Metadata about entities and relationships | | state | object | Canonical world state snapshot with integrity hash | | dynamics | object | Transition rules and causal structure | | uncertainty | object | Confidence levels and unknown factors | | provenance | object | Source tracking and lineage | | simulation | object | Model validation results | | summary | object | Human-readable insights | | confidence | number | 0.0-1.0 based on evidence completeness | | evidence_anchors | array | All evidence references collected | | assumptions | array | Explicit assumptions made during modeling |

Examples

Example 1: Build Supply Chain World Model

Input:

goal: "Model electronics supply chain for disruption risk analysis"
scope:
  - "suppliers"
  - "manufacturers"
  - "logistics"
  - "inventory"
constraints:
  time_horizon: "6 months"
  geographic_scope: "Asia-Pacific"
  confidence_threshold: 0.7
sources:
  - type: database
    connection: "postgres://supply-chain-db"
  - type: api
    endpoint: "https://logistics.api/shipments"

Output:

workflow_id: "world_20240115_100000_supplychain"
goal: "Model electronics supply chain for disruption risk analysis"
status: completed
world_model:
  version: "v1.0.0"
  created_at: "2024-01-15T10:00:00Z"
  schema_version: "world_state_schema_v2"
  entities:
    count: 1247
    by_type:
      supplier: 156
      manufacturer: 23
      warehouse: 45
      distribution_center: 12
      product: 892
      shipment: 119
    sample:
      - id: "supplier-taiwan-001"
        type: "supplier"
        name: "Taiwan Semiconductor Co"
        location: "Hsinchu, Taiwan"
        capacity: 50000
        lead_time_days: 45
      - id: "mfg-shenzhen-005"
        type: "manufacturer"
        name: "Shenzhen Electronics Assembly"
        location: "Shenzhen, China"
        capacity: 100000
  relationships:
    count: 3456
    types:
      - "supplies_to"
      - "located_in"
      - "transports_via"
      - "stores_at"
      - "depends_on"
    sample:
      - subject: "supplier-taiwan-001"
        predicate: "supplies_to"
        object: "mfg-shenzhen-005"
        attributes:
          volume: 25000
          frequency: "weekly"
  evidence_anchors:
    - "tool:database:supply-chain-db/entities"
    - "tool:api:logistics.api/shipments"
state:
  snapshot:
    timestamp: "2024-01-15T10:00:00Z"
    entities: "[1247 entities - see world_state.yaml]"
    relationships: "[3456 relationships - see world_state.yaml]"
  hash: "sha256:def456abc789..."
  timestamp: "2024-01-15T10:00:00Z"
  evidence_anchors:
    - "file:state/supply_chain_world.yaml"
dynamics:
  transition_rules: 34
  causal_links: 89
  temporal_scope: "daily"
  evidence_anchors:
    - "tool:state-transition:rule_extraction"
    - "tool:causal-model:dependency_graph"
uncertainty:
  overall_confidence: 0.82
  high_uncertainty_areas:
    - "Supplier capacity utilization (estimated from public data)"
    - "Shipping delays (historical average, not real-time)"
  unknown_factors:
    - "Competitor orders affecting supplier allocation"
    - "Regulatory changes in transit countries"
  evidence_anchors:
    - "tool:uncertainty-model:confidence_analysis"
provenance:
  sources:
    - "postgres://supply-chain-db (primary)"
    - "https://logistics.api (secondary)"
    - "public filings (supplementary)"
  lineage_depth: 3
  coverage: 0.94
  evidence_anchors:
    - "tool:provenance:lineage_trace"
simulation:
  validated: true
  scenarios_tested: 5
  anomalies_found:
    - "Taiwan supplier shutdown causes 67% production halt within 2 weeks"
    - "Shipping route disruption adds 12-day average delay"
  evidence_anchors:
    - "tool:simulation:scenario_results"
summary:
  description: "Electronics supply chain model covering 156 suppliers, 23 manufacturers, and supporting logistics infrastructure in Asia-Pacific region"
  key_insights:
    - "Single-source dependency on Taiwan for 45% of semiconductor supply"
    - "Shenzhen manufacturing hub handles 60% of assembly volume"
    - "Average supply chain depth of 3 tiers with limited visibility beyond tier 1"
  recommended_actions:
    - "Diversify semiconductor sourcing to reduce Taiwan concentration risk"
    - "Establish buffer inventory for critical components"
    - "Develop secondary logistics routes for key shipping lanes"
  evidence_anchors:
    - "tool:summarize:executive_summary"
confidence: 0.82
evidence_anchors:
  - "tool:database:supply-chain-db"
  - "tool:api:logistics.api"
  - "tool:simulation:scenario_results"
  - "file:state/supply_chain_world.yaml"
assumptions:
  - "Database reflects current operational state"
  - "API provides accurate shipment tracking"
  - "Public capacity data is within 20% of actual"
  - "Lead times based on historical 90-day average"

Evidence pattern: Multi-source data integration, entity resolution across databases, causal analysis from transaction patterns, uncertainty from data freshness and coverage.

Verification

[ ] Entity Coverage: All entities in scope identified with canonical IDs
[ ] Relationship Completeness: Key relationships mapped with evidence
[ ] State Validity: World state conforms to schema
[ ] Dynamics Defined: Transition rules and causal links documented
[ ] Uncertainty Quantified: Confidence scores for all major assertions
[ ] Provenance Complete: Source documented for >90% of facts
[ ] Simulation Validated: At least 1 scenario successfully executed

Verification tools: Read (for state files), Bash (for simulation), Web (for API validation)

Safety Constraints

mutation: false
requires_checkpoint: false
requires_approval: false
risk: medium

Capability-specific rules:

Do not modify source data during modeling
Flag entities with confidence < threshold
Document all assumptions explicitly
Preserve raw data alongside derived state
Validate schema conformance before completion
Rate-limit API calls to respect source limits

Composition Patterns

Commonly follows:

retrieve - After gathering raw data
receive - After ingesting real-time signals
inspect - After initial data quality assessment

Commonly precedes:

digital-twin-sync-workflow - World model is prerequisite for sync
simulate - To run what-if scenarios
forecast-risk - To predict future states
summarize - To generate executive reports

Anti-patterns:

Never skip identity resolution before state construction
Never omit uncertainty modeling for production use
Never finalize without provenance documentation
Never deploy model without simulation validation

Workflow references:

See reference/workflow_catalog.yaml#world-model-workflow for step definitions
See reference/world_state_schema.yaml for canonical state format