Validate field atlas extractions

Independently deep-read a sample and check FieldAtlas

Environment or artifact verification with mechanical pass/fail detection against ground truth.

• · args: configuration object
• · file_content (markdown/text)
• · grep_results (text)

• · markdown_files (md)
• · structured_data (JSON)
• · verdict_reports (JSON)

• · agent
• · phase
• · pipeline
• · bash
• · file_read
• · sqlite
• · json_schema

• · Specific repo paths or tool locations configured
• · Python virtual environment or package manager set up
• · Workflow orchestrator runtime context
• · LLM agent tier models available
• · Async runtime environment

• · Agent timeout or non-response
• · Phase execution state corruption
• · Command execution failure or non-zero exit
• · Promise rejection or timeout
• · Database query or connection failure
• · Malformed input data
• · Empty result set causing downstream failures

• · Structured output schemas with required-field validation
• · Explicit enumeration of findings with severity/category classification
• · Ground-truth validation against authoritative source
• · Iteration history tracked for convergence analysis
• · Explicit specification or contract documentation