# Image Upload Preflight Standalone local batch header inspector for PNG, JPEG and WebP. Python 3.10+ standard library; Tkinter required for native UI. No image packages, account or other purchase. Run `python app.py`, load input JSON, inspect and save a new report. CLI: `python checker.py input.json new-report.json`. Generate a real example PNG with `python examples/make_sample.py` and use examples/input.json. Absolute file paths recommended; native UI resolves relative paths from this product folder. Existing output reports are refused; source files are not modified. Supply files (1–100 paths), optional limits max_bytes, min_edge, max_edge, max_pixels. Defaults: 12MiB, minimum 128px, maximum 8192px, 24 million pixels. Hard ceilings: 32MiB/file, 16384px/edge, 100 million pixels. Sequential bounded reads retain at most one input image plus small reports. Animated APNG/WebP always refused. Unsupported GIF, HEIC, AVIF, SVG and PDF are explicit per-file refusals. Missing files also remain in the report. No network, image upload or full pixel decoding. JSON contains headerAccepted for each file, dimensions/format, file hash when fully read, refusal codes and one-buffer RGBA estimate. That estimate is width*height*4, not measured decoder RAM. A report saved successfully may contain rejected files; check summary/refused and individual rows. Corruption past inspected headers can still exist: this is not full decoding, PNG CRC verification, content moderation, EXIF stripping or image repair. JPEG orientation is not applied. PNG chunk lengths are bounded through the first IDAT, and WebP RIFF size is checked. This closes inherited acceptance of truncated PNG chunks and scanning only the first 64 chunks before animation metadata. Actual rendering remains the decoder's responsibility. Tests: `python -m unittest discover -s tests -v`. Native visual acceptance is separate and pending.