goals-extract

name	goals-extract
description	Extract business goals, success criteria, constraints, KPIs, assumptions, and open questions from requirements input.
license	MIT

Goals Extract Skill

Intent

Extract outcome goals from requirements using rule-based pattern matching and LLM-based extraction. Focus specifically on measurable business outcomes rather than system mandates or technical constraints.

Inputs

• Source: projects/[project-name]/artifacts/Analysis/requirements.json (generated by requirements-ingest skill)
• Format: Structured requirements data with IDs and sections for traceability

Outputs

Files Generated:

• projects/[project-name]/artifacts/Analysis/goals.json - Structured data for programmatic use
• projects/[project-name]/artifacts/Analysis/goals.md - Human-readable documentation

JSON Structure (goals.json)

{
  "goal_statement": "string",
  "success_criteria": [
    {
      "criteria": "string",
      "source_refs": ["req_id:section"]
    }
  ],
  "kpis": [
    {
      "kpi": "string", 
      "source_refs": ["req_id:section"]
    }
  ],
  "constraints": [
    {
      "constraint": "string",
      "source_refs": ["req_id:section"]
    }
  ],
  "assumptions": [
    {
      "assumption": "string",
      "source_refs": ["req_id:section"]
    }
  ],
  "open_questions": [
    {
      "question": "string",
      "source_refs": ["req_id:section"]
    }
  ],
  "traceability": {
    "source_requirements": ["req_id"],
    "extraction_timestamp": "ISO8601",
    "confidence_scores": {
      "goal_statement": "0.0-1.0",
      "overall": "0.0-1.0"
    }
  }
}

Markdown Structure (goals.md)

# Goals Analysis

## Business Goal
[Goal statement with inline reference links]

## Success Criteria
- [Criteria 1] *(Ref: req_id:section)*
- [Criteria 2] *(Ref: req_id:section)*

## Key Performance Indicators
- [KPI 1] *(Ref: req_id:section)*
- [KPI 2] *(Ref: req_id:section)*

## Constraints
- [Constraint 1] *(Ref: req_id:section)*
- [Constraint 2] *(Ref: req_id:section)*

## Assumptions
- [Assumption 1] *(Ref: req_id:section)*
- [Assumption 2] *(Ref: req_id:section)*

## Open Questions
- [Question 1] *(Ref: req_id:section)*
- [Question 2] *(Ref: req_id:section)*

---
**Traceability:** Extracted from requirements: [req_id_1], [req_id_2]  
**Generated:** [timestamp]

Tools

Rule-based Pattern Extraction (Primary)

• spaCy/Universal Dependencies for dependency parsing and grammatical analysis
• Outcome cue patterns: increase|reduce|improve|decrease|optimize|ensure|enhance|streamline|grow|expand
• Intent signal patterns: "should aim to", "objective is to", "goal is to", "target is", "we will", "by [date]", "within [timeframe]"
• Keyphrase extraction: RAKE/YAKE/TextRank constrained to outcome-containing sentences
• OpenIE extraction: Get (subject, relation, object) tuples where relation contains outcome verbs
• Negative filtering: Exclude "the system shall..." unless followed by outcome language

LLM-based Extraction (Secondary)

• LLM with outcome-focused extraction prompt
• Confidence scoring and validation

Supporting Tools

• JSON parser to read requirements.json from Analysis folder
• Glossary lookup to normalize terms
• File writer for dual output format (JSON + Markdown)

Extraction Algorithm

Phase 1: Rule-based Pattern Matching

1. Outcome Pattern Detection:

   • Scan for outcome cue words: increase|reduce|improve|decrease|optimize|ensure|enhance|streamline|grow|expand
   • Identify modal/intent signals: "should aim to", "objective is to", "goal is to", "target is", "we will", "by [date]", "within [timeframe]"

2. Dependency Analysis (spaCy/UD):

   • Extract (VERB→dobj) patterns where verb ∈ outcome verbs
   • Expand noun phrases and capture adjuncts (prep: by/within/to) for targets/deadlines

3. Negative Filtering:

   • Exclude system mandates: "the system shall..." unless followed by outcome language ("to reduce...")
   • Filter out pure technical constraints

4. Keyphrase Extraction:

   • Apply RAKE/YAKE/TextRank to sentences containing outcome patterns
   • Constrain to outcome-relevant phrases for precision

5. OpenIE Extraction:

   • Extract (subject, relation, object) tuples
   • Keep tuples where relation contains outcome verbs

Phase 2: LLM Validation & Enhancement

• Validate pattern-extracted goals for coherence
• Fill gaps in goal statements
• Generate confidence scores
• Ensure SMART criteria compliance for measurable outcomes

Quality Criteria

• Outcome Focus: Goals must describe desired business outcomes, not system features
• SMART Compliance: Specific, Measurable, Achievable, Relevant, Time-bound
• Pattern Precision: Rule-based extraction achieves >90% precision on outcome identification
• Traceability: Every extracted goal links to source requirement with confidence score
• Negative Filtering: Successfully excludes technical mandates and constraints

Example Output

JSON Format

{
  "goal_statement": "Reduce requirements processing time by 75% within 6 months while maintaining 99% accuracy.",
  "success_criteria": [
    {
      "criteria": "Achieve 75% reduction in processing time compared to current baseline.",
      "source_refs": ["REQ-001:2.1"],
      "extraction_method": "pattern_match",
      "outcome_verb": "reduce"
    },
    {
      "criteria": "Maintain 99% accuracy rate for all processed requirements.",
      "source_refs": ["REQ-002:3.1"],
      "extraction_method": "pattern_match", 
      "outcome_verb": "maintain"
    },
    {
      "criteria": "Complete implementation within 6 months of project start."
       
       
       
    
  
   
    
       
       
       
       
       
       
    
    
       
       
       
       
       
       
    
  
   
    
       
       
    
    
       
       
    
    
       
       
    
  
   
    
       
       
    
    
       
       
    
    
       
       
    
  
   
    
       
       
    
    
       
       
    
    
       
       
    
  
   
        
     
     
       
       
       
       
    
  

Markdown Format

# Goals Analysis

## Business Goal
Reduce requirements processing time by 75% within 6 months while maintaining 99% accuracy.

## Success Criteria
- Achieve 75% reduction in processing time compared to current baseline. *(Ref: REQ-001:2.1) [reduce]*
- Maintain 99% accuracy rate for all processed requirements. *(Ref: REQ-002:3.1) [maintain]*
- Complete implementation within 6 months of project start. *(Ref: REQ-001:4.2) [complete]*

## Key Performance Indicators
- Processing time: < 2.5 minutes per batch (75% reduction from 10 minutes baseline) *(Ref: REQ-002:2.3) [reduce]*
- Accuracy rate: ≥ 99% (unit: percent, timeframe: daily) *(Ref: REQ-002:3.1) [maintain]*

## Constraints
- Input requirements must conform to the specified format. *(Ref: REQ-001:1.1)*
- System must operate within existing infrastructure and security policies. *(Ref: REQ-004:1.1)*
- No manual intervention allowed during ingestion. *(Ref: REQ-002:2.5)*

## Assumptions
- Requirements provided are complete and up-to-date. *(Ref: REQ-001:assumptions)*
- Glossary terms are available for normalization. *(Ref: REQ-003:2.1)*
- Downstream systems are ready to consume ingested data. *(Ref: REQ-004:3.2)*

## Open Questions
- What is the process for handling ambiguous or conflicting requirements? *(Ref: REQ-001:notes)*
- How will updates to requirements be managed post-ingestion? *(Ref: REQ-002:future_considerations)*
- Are there scalability limits for batch size or frequency? *(Ref: REQ-002:performance)*

---
 Extracted from requirements: REQ-001, REQ-002, REQ-003, REQ-004  
 2026-02-08T10:30:00Z  
 Rule-based patterns + LLM validation  
 0.92 |  0.89 |  0.87