Security Policy, Emerging Cyber Threats, Research & AI
SPECTRA is an AI-automated monthly cybersecurity policy digest. It collects from 17 public government and industry sources, uses Claude to categorize and summarize items, and renders a branded PDF report -- all in a single scheduled task run.
Built for DoD/federal cybersecurity practitioners who need to stay current on policy, threats, and AI developments without manually scanning dozens of feeds.
A monthly PDF digest with seven sections:
| Section | What's Covered |
|---|---|
| Executive Summary | Top 5 items by relevance |
| Policy & Compliance | Agency actions, EOs, CISA directives, FedRAMP, CMMC |
| Publications & Standards | NIST SPs, CISA guides, strategy docs |
| Threats & Incidents | Breaches, APT campaigns, KEVs, ICS advisories |
| AI & Agentic Developments | Frontier models, AI policy, agentic capabilities |
| Legislative Highlights | Bills with significant movement only |
| Upcoming Conferences | 5-10 events for the next 2-3 months |
Each section is limited to the top 20 items by relevance score. Items are scored 1-10 based on direct impact to federal cyber practitioners.
Government (8): NIST CSRC, CISA Alerts, CISA KEV Catalog, Federal Register API, Congress.gov API, Executive Orders, FedRAMP Notices, FedRAMP Changelog
Industry News (4): CyberScoop, Nextgov/FCW, Dark Reading, The Record
AI (5): OpenAI Blog, Google AI Blog, DeepMind Blog, The Verge AI, MIT Technology Review AI
All sources are public and unclassified. See src/collect/config.yaml for the full configuration.
The entire pipeline runs as a single Claude Code scheduled task. Claude reads the raw items, categorizes each into the appropriate section, writes a 2-3 sentence summary, scores relevance, consolidates duplicate stories, and assembles the markdown draft. No API key needed -- runs on the Claude Code subscription.
- Python 3.10+
- Claude Code (for the scheduled curation task)
git clone https://github.com/JAKSecurity/SPECTRA.git
cd SPECTRA
pip install -r requirements.txt# 1. Collect from all sources
PYTHONPATH="." python -m src.collect.runner src/collect/config.yaml data/sources
# 2. Prep (load + dedup)
PYTHONPATH="." python -m src.curate.curate data/sources/2026-04 data/drafts/prepped_2026-04.json
# 3. Curate -- this step requires Claude Code (scheduled task handles it)
# See the scheduled task SKILL.md for the full curation prompt
# 4. Render to PDF
PYTHONPATH="." python -m src.render.render data/drafts/2026-04-SPECTRA.md output/2026-04-SPECTRA.pdfThe spectra-monthly Claude Code scheduled task runs the full pipeline at 5AM on the 1st of each month. It collects sources, curates with AI, presents the draft for review, renders the PDF, and delivers via email and Discord.
python -m pytest tests/ -q59 tests covering fetchers, prep/dedup, draft assembly, and PDF rendering.
src/
collect/ Source fetcher modules and config
config.yaml Source URLs, types, category hints
runner.py Orchestrates fetchers, writes health report
fetchers/ One module per source type (RSS, Federal Register API, etc.)
curate/ Item loading, dedup, and draft assembly
render/ PDF template and rendering (reportlab)
scripts/ Shell entry points for each stage
tests/ 59 tests (pytest)
docs/ Design spec and backlog