Decision Orchestrator

Data-Driven Decision Orchestrator

Core Principles

Data drives decisions, but decision authority belongs to humans

The role of data is to illuminate blind spots in decisions, not to replace decision-makers. In the DACE loop, Define and Analyze are data-driven, Conclude is human-decided, Execute is system-tracked -- this is the optimal division of labor between data and humans.

Orchestration Philosophy

1. DACE loop is the main thread, insights embedded, culture is the support: The DACE loop drives the decision closed loop; the Analyze stage has integrated insight conversion capability; culture building ensures decisions are implemented
2. Conclude stage must have human participation: No matter how clear the data, decisions involving business strategy must be confirmed by humans
3. Decision boundary tiered delegation: datadecision auto-executed, datareference pushed for confirmation, human_decision waits for approval

Orchestration Protocol

The orchestration protocol follows the unified standard in [orchestrator-protocol.md](../../../codex-templates/orchestrator-protocol.md).

Pipeline

pipeline: decision-orchestrator
version: 7.0
post_pipeline:
  - action: stage-summary
    output: output/phase-reports/pm-metrics-ops/decision-orchestrator.md

stages:
  - id: phase-1
    name: "DACE Decision Loop"
    depends_on: []
    skills: [decision-dace]
    gate:
      condition: "Objectives defined, data analyzed, insights generated, decision options provided"
      fail_action: "Supplement data or redefine objectives"

  - id: phase-2
    name: "Data Culture Building"
    depends_on: [phase-1]
    skills: [decision-culture]
    gate:
      condition: "Report system operating normally (daily/weekly/monthly/quarterly)"
      fail_action: "Check upstream data sources or adjust report templates"

Stage Execution Plan

Invoke decision-dace

Invoke: ${decision-dace}
Input:
  okr_data: provided by user
  kr_progress: analysis-anomaly -> anomaly_report.json
  experiment_result: experiment-execution -> experiment_result.json
  analysis_result: analysis-anomaly -> anomaly_report.json
  business_context: provided by user (optional)
  insight_library: decision-dace -> insight_library.json (optional)
Output: output/pm-metrics-ops/decision-dace/
Validation: Define stage objectives quantifiable with baselines; Analyze stage covers all data sources; Conclude stage provides at least 2 decision options; Execute stage sets monitoring and rollback mechanisms; insight narratives use business language not data jargon; each insight provides at least 2 decision options; decision boundaries labeled correctly (auto/reference/human); recommended actions have clear next steps and owners
Mode: AI->Human

Invoke decision-culture

Invoke: ${decision-culture}
Input:
  okr_data: decision-dace -> dace_status.json
  decision_records: decision-dace -> decision_insight.json
  team_feedback: provided by user (optional)
Output: output/pm-metrics-ops/decision-culture/
Validation: Daily summary produces no noise alerts when no anomalies; weekly report includes OKR progress and experiment summary; monthly report includes complete metric trends and deviation analysis; all data references in reports traceable to data sources
Mode: AI->Human

Stage Summary (post_pipeline)

Follows the stage summary protocol in [orchestrator-protocol.md](../../../codex-templates/orchestrator-protocol.md).

| Parameter | Value | |------|-----| | Sub-Skill output path | output/pm-metrics-ops/ | | Summary output path | output/phase-reports/pm-metrics-ops/decision-orchestrator.md |

Downstream connections: primary: design-orchestrator (decision complete, convert decision conclusions into feature changes) alternatives:

• target: experiment-orchestrator

reason: Decision needs A/B testing to validate effectiveness condition: Decision conclusions need quantitative verification

• target: iteration-orchestrator

reason: Decision involves iteration priority adjustment condition: Decision conclusions impact iteration plan special_cases:

• target: decision-dace

reason: Only need DACE decision loop, no full decision orchestration required condition: Already have analysis conclusions, only need quick decision closed loop

Stage Gates

| Gate | Condition | Failure Handling | |------|------|------------| | DACE loop Define/Analyze complete | dace-define-analyze output file generated and non-empty | Supplement data or redefine objectives | | Decision options provided | decision-options output file generated and non-empty | Mark as pending, continuously track | | Data culture report system running | data-culture-report output file generated and non-empty | Check upstream data sources or adjust report templates | | Stage summary generated | output/phase-reports/pm-metrics-ops/decision-orchestrator.md generated and all 6 structural items non-empty | Supplement missing structural items and regenerate |

Human Decision Points

| Decision Point | Trigger Condition | Decision Content | |--------|----------|----------| | Conclude stage decision | DACE loop enters Conclude stage | Review analysis conclusions, make final decision |

Decision Boundary Management

| Decision Type | Description | Execution Method | |---------|------|---------| | datadecision | Data clearly supports, can auto-execute | AI auto-executes + post-hoc report | | datareference | Data for reference, human decides | Push insights, wait for decision | | human_decision | Complex decision, human-led | Provide analysis, human decides |

Exception Handling

| Exception Type | Handling Strategy | |----------|----------| | DACE loop Conclude stage human not responding | Pause Execute stage, preserve Conclude state, support resuming after human returns | | Insight confidence too low ( Produces PRD Continue with AI-inferred values? (Inferred values confidence =0.5 | When confidence project_dir validity -> Input confidence assessment

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: LuckyOneTwoThree
• Source: LuckyOneTwoThree/vibe-skill
• License: MIT
• Homepage: https://luckyonetwothree.github.io/all-skill-html/

Install and usage instructions live in the source repository linked above.