regen-reports

name	regen-reports
description	Regenerate the markdown and diagram report files for one or all corpus architectures from their existing ground_truth.json and MoE JSON files, without re-running the analysis engine or Expert Review. Use after code changes to threat_report.py, improvement_summary_generator.py, executive_dashboard_generator.py, adr_generator.py, or narrative_enricher.py to keep all report files current. Also supports regenerating only specific report layers (md-only, dashboard-only, improvement-only).
allowed-tools	Bash(python3:*) Bash(source:*)

regen-reports — Report File Regeneration

Regenerates report .md and .mmd files from existing JSON data (ground_truth.json + MoE/SM JSON files). No engine re-run, no LLM calls, no Expert Review required.

Use this after changes to any report generator module to propagate the fix across the full corpus.

Run

# Regenerate all report files for one architecture (MD + diagrams + dashboard + improvement)
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py 21_agentic_ai_system

# Regenerate all 26 corpus architectures
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py --all

# Regenerate only the markdown reports (01-03, 09, 10) — fastest
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py 21_agentic_ai_system --md-only

# Regenerate only 00_executive_dashboard.md (requires 07_moe_orchestrator.json)
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py 21_agentic_ai_system --dashboard-only

# Regenerate only 08_improvement_summary.md (requires 07_moe_orchestrator.json)
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py 21_agentic_ai_system --improvement-only

# Dry-run: show what would be regenerated without writing files
cd "$(git rev-parse --show-toplevel)" && source .venv/bin/activate \
  && python3 .claude/skills/regen-reports/scripts/regen-reports.py --all --dry-run

What gets regenerated

Layer	Files	Requires	When to use
MD reports	01_executive_summary.md, 02_technical_report.md, 03_action_plan.md, 09_threat_model.md, 10_adr_report.md	ground_truth.json	After changes to threat_report.py
Diagrams	before.mmd, after.mmd, threatmodel_adr.mmd, threatmodel_adr_bh.mmd	ground_truth.json	After changes to diagram generators
Dashboard	00_executive_dashboard.md	07_moe_orchestrator.json	After changes to executive_dashboard_generator.py
Improvement	08_improvement_summary.md, 08a/b/c_*.mmd	07_moe_orchestrator.json	After changes to improvement_summary_generator.py

Layers without their required JSON files are silently skipped with a note in the output.

Output format

Regen Reports — 21_agentic_ai_system

  ✓  01_executive_summary.md
  ✓  02_technical_report.md
  ✓  03_action_plan.md
  ✓  09_threat_model.md
  ✓  10_adr_report.md
  ✓  before.mmd
  ✓  after.mmd
  ✓  threatmodel_adr.mmd
  ✓  00_executive_dashboard.md  [requires 07_moe_orchestrator.json]
  ✓  08_improvement_summary.md  [requires 07_moe_orchestrator.json]
  –  08a_quick_wins.mmd          [skipped — no improvement_options in MoE data]

  Done: 10 regenerated · 1 skipped · 0 failed  (2.4s)

When to run

• After any change to chatbot/modules/threat_report.py
• After any change to chatbot/modules/improvement_summary_generator.py
• After any change to chatbot/modules/executive_dashboard_generator.py
• After any change to chatbot/modules/adr_generator.py or narrative_enricher.py
• After upgrading ATLAS or MITRE data (technique names/descriptions change)
• Before sharing or reviewing a report that may be stale

Notes

• Does NOT re-run the analysis engine or Expert Review — existing JSON data is the source of truth
• For each architecture, the script loads ground_truth.json and any present MoE/SM JSON files
• Architectures missing ground_truth.json are skipped
• --all processes all directories under report/ with a ground_truth.json
• Takes ~2–5 seconds per architecture (pure Python, no LLM calls)

Related skills

• /rerun-corpus — re-run the deterministic analysis engine (rewrites ground_truth.json)
• /run-er — run Expert Review to update MoE/SM JSON files
• /tatb-score — verify TATB scores after regenerating reports