UAP Declassified Release Analyzer

Problem

Declassified UAP/UFO releases usually arrive as a folder of PDFs: messy filenames, mixed agencies, many scanned files without text layers, and FOIA redactions scattered across documents. Manual review only gives fragmented impressions, making it hard to answer which files are most significant, which agencies or locations recur, or which tranche is most redacted.

How It Works

The skill packages a repeatable pipeline: inventory.py builds a file inventory and classifies sources such as FBI, DOW, NASA, DOS, and NARA by filename prefix; extract_text.py pulls text from PDFs that have a text layer and marks scanned files as OCR-needed; analyze.py surfaces entities, year clusters, agency-by-location patterns, frequent names, and FOIA codes; build_report.py writes REPORT.md. The workflow is idempotent and incremental, so reruns on the same folder skip completed work.

Limits and Notes

It is intended for already downloaded FOIA release folders and is useful for inventories, summaries, and pattern surfacing. It does not run OCR by default and does not treat year mentions as incident dates. If many filenames fall into OTHER, extend agency_vocab.md instead of assuming the data is malformed. Entity extraction is keyword- and regex-based, so the report should keep its “missing data and caveats” section and avoid presenting heuristic results as formal NER.