Coverage
Coverage and freshness
We collect a small set of approved public sources on their own schedules, separately from queries. Collection targets aren't indexing guarantees. There's no general web crawling, no OCR and no generated answers.
| Source | Content | Update target |
|---|---|---|
| CISA Known Exploited Vulnerabilities | Full catalog records | Frequent polling |
| Defense.gov news | Official RSS excerpts | Frequent polling |
| Federal Register (defense acquisition) | Document text | Daily |
| eCFR (acquisition security) | Regulation sections | Daily |
| arXiv (AI and security) | Metadata and abstracts | Daily |
| National Weather Service alerts (Mid-Atlantic) | Alert text (expires) | Frequent polling |
| NIST publications | Reviewed full-text passages with page citations | Reviewed snapshots |
| NIST OSCAL | Structured control catalogs | Reviewed snapshots |
Agents see the same information through GET /v1/sources, and every result carries checked_at and stale. The next source pack is chosen from evaluation tasks.