Agent Skills: Dark Intelligence Workflow

Dark Intelligence Workflow

Overview

Use this skill when porting engineering calculations from legacy Excel spreadsheets (or other legacy formats) into clean, client-free implementations in the 4 public repos. The workflow extracts only generic methodology — equations, input/output schemas, worked examples — and strips all client/project identifiers before archival.

Sub-Skills

• Inputs
• Step 1 — Identify (+4)
• Step 6 — Implement (+1)
• Integration Points
• AC Checklist

Ecosystem Audit & Promotion Planning

Use this workflow when you need to assess the full dark intelligence landscape across the ecosystem and plan promotion work into GitHub issues.

3-Parallel-Subagent Audit Pattern

Run 3 subagents simultaneously to cover the full surface:

1. Excel Inventory Subagent — scan for:

   • Actual .xlsx/.xls files on disk and in repo
   • Dark intelligence extraction outputs (knowledge/dark-intelligence/)
   • Excel-processing scripts (scripts/data/doc_intelligence/)
   • Catalog/index entries referencing Excel (data/document-index/, docs/CONTENT_INDEX.md)
   • Planning docs (.planning/algorithm-extraction.md, heavy-batch-prompts.md)

2. Conversion Infrastructure Subagent — map:

   • Skills: dark-intelligence-workflow, xlsx-to-python, doc-extraction, doc-intelligence-promotion
   • Pipeline scripts: parsers, formula_to_python, chain builder, pattern detector, module assembler
   • POC outputs: xlsx-poc/ (v1 formula extraction), xlsx-poc-v2/ (pattern + code gen)
   • digitalmodel Excel utilities (legacy/ scripts, excel_utilities.py modules)
   • Schema & validation (config/schemas/dark-intelligence-archive.yaml)

3. Domain Coverage Subagent — assess:

   • Module registries (solver registry, design code registry, skill graph index)
   • Standards transfer ledger (data/document-index/standards-transfer-ledger.yaml)
   • Domain file counts in digitalmodel/src/digitalmodel/ (30 domains, 1,362 files)
   • Function counts and standards-mapped vs unmapped (docs/vision/CALCULATIONS-VISION.md)
   • Test coverage baselines (config/testing/coverage-baseline.yaml)

Gap Matrix Output

Cross-reference the 3 subagent outputs into a promotion map:

EXCEL SOURCE          → EXTRACTION STATE   → CODE COVERAGE    → STANDARDS STATUS
/mnt/ace/*.xls (3.6K)   6 extracted          1,362 files        29/425 done
CONTENT_INDEX (419)      0 extracted          7,355 functions    235 gaps
Standards DB             cataloged            2.95% test cov     93 materials gaps

Issue Generation Pattern

Structure GitHub issues as:

• 1 epic with child issue cross-references
• Phase 1: Wire existing extractions (quick wins — extracted but not promoted)
• Phase 2: Batch extraction (high-value unprocessed Excel files)
• Phase 3: Systematic gap closure (by domain, largest gaps first)
• Phase 4: Infrastructure (test coverage uplift, feedback loop, registry rebuild)

Labels: dark-intelligence, domain:code-promotion, domain:extraction-pipeline, cat:engineering-calculations

Key Reference Files

• Assessment output: docs/document-intelligence/dark-intelligence-excel-assessment.md
• Standards ledger: data/document-index/standards-transfer-ledger.yaml
• Domain coverage: docs/document-intelligence/domain-coverage.md
• Calculations vision: docs/vision/CALCULATIONS-VISION.md
• Pipeline scripts: scripts/data/doc_intelligence/
• Dark intel archives: knowledge/dark-intelligence/
• Algorithm plan: .planning/algorithm-extraction.md