Phase 1: Ingest

Read and catalog all input materials to prepare for analysis.

Trigger

Execute when:

• Starting Ontology Builder Pipeline
• New documents added to _input/
• Re-processing requested

Process

Step 1: Validate Input Structure

Check _input/ folder:

Required:
✓ project-context.md (MUST exist)

Optional but valuable:
○ domain-hints.md
○ requirements/*.md
○ interviews/*.md
○ existing-docs/*
○ references/*

If project-context.md missing: → STOP and request from user

Step 2: Read Project Context

Extract from project-context.md:

• Project name, module, domain
• Region (for regulatory context)
• Current state and pain points
• Stakeholders
• Constraints

Store as context_vars for later phases.

Step 3: Catalog All Files

For each file in _input/:

file_entry:
  path: [relative path]
  type: [user-story | interview | existing-doc | reference | other]
  format: [md | pdf | docx | txt | image]
  size: [file size]
  summary: [2-3 sentence summary of content]
  key_topics: [list of main topics]
  entities_mentioned: [list of nouns that might be entities]
  workflows_mentioned: [list of processes/actions mentioned]

Step 4: Identify File Relationships

Map relationships between files:

• Which files reference each other?
• Which files cover same topics?
• Which files contradict each other?

Step 5: Assess Input Quality

Rate input completeness:

Step 6: Generate Ingestion Report

Output: _output/_logs/ingestion-report.md

# Ingestion Report

**Generated**: [timestamp]
**Project**: [from context]
**Module**: [from context]

## Input Summary

| Category | Files | Status |
|----------|-------|--------|
| Project Context | 1 | ✓ Valid |
| Requirements | [N] | [status] |
| Interviews | [N] | [status] |
| Existing Docs | [N] | [status] |
| References | [N] | [status] |

## File Catalog

[List all files with summaries]

## Preliminary Extraction

### Potential Entities (Raw)
[List all nouns that appear to be domain entities]

### Potential Workflows (Raw)
[List all processes/actions mentioned]

### Potential Business Rules (Raw)
[List any rules or constraints mentioned]

## Quality Assessment

[Quality scores and notes]

## Gaps Identified

[List what seems to be missing]

## Recommendations

[Suggestions for Phase 2]

Output

• _output/_logs/ingestion-report.md
• _output/_logs/gate-1-manifest.yaml (verification manifest)
• Ready for Phase 2: Analyze

Gate 1: Self-Verification

Before completing Phase 1, generate verification manifest:

# _output/_logs/gate-1-manifest.yaml
gate: 1
name: "Post-Ingest Verification"
timestamp: "[ISO timestamp]"

structural_checks:
  - check: "ingestion-report.md exists"
    status: PASS
  - check: "All input files cataloged"
    status: PASS | FAIL
    details:
      input_files_count: [N]
      cataloged_count: [N]
      missing: []
  - check: "project-context.md processed"
    status: PASS | FAIL

consistency_checks:
  - check: "No duplicate entries"
    status: PASS | FAIL
  - check: "All file paths valid"
    status: PASS | FAIL

traceability_checks:
  - check: "Each file has summary"
    status: PASS | FAIL
    files_without_summary: []

result:
  status: PASS | FAIL | WARN
  blocking_failures: []
  warnings: []
  proceed_to_next_phase: true | false

Verification Rule: Only proceed to Phase 2 if status: PASS or status: WARN

Error Handling

Next Phase

After completing ingestion: → Load phase-2-analyze/SKILL.md → Pass ingestion-report.md as input