Receipt Chain Inspector

TL;DR: paste URL or attach a file, pick Intended use, auto-apply musts, Archive, export MD/JSON.

Short version:

• URL + Type + Intended use = what courtroom you’re walking into.
• Claims = what the artifact is being asked to carry.
• Checks = what receipts you already have.
• (Optional) File = attach a hash + basic metadata.
• Score = how expensive it is to fake this, not whether it’s true.

What each field actually does

• Original URL / description – Starting point for archiving and linking; goes into the share-safe summary.
• Artifact type – Labels the chain; also nudges what “musts” look like (see below).
• Intended use
  • Internal decision → lighter; team receipt is fine.
  • Public post → requires redaction log + archived source.
  • Press / legal → you’re in evidence land; all musts apply.
  • Private share → fewer; but still hash and context.
• Claims (one per line) – Defines the burden. More claims ⇒ longer chain.
• Optional checks (tick what you already have)
  • Signed at capture (C2PA or eq.)
  • Source still reachable (archived)
  • Hash of original file recorded
  • Timestamp & operator recorded
  • Context captured (surrounding UI/thread)
  • Share-safe redaction made (crop/blur log)
  • Public receipt link posted
    Ticking these both raises the score and moves items from “need” to “have” in the output.
• Optional file (image/video/PDF) – Tool computes SHA-256, size, and basic media metadata and injects it into the Markdown (“Attached file” block).

Chain strength score (how it’s computed)

• Simple: (checks ticked / total checks) * 100, rounded.
• It’s a cost-to-fake proxy, not a truth meter. In press/legal, aim for ~100. For public posts, >70 is sane. Under 40? That’s a rumor.

“Musts” by artifact type (practical extras)

• Image → prefer C2PA; if not, hash + full-frame + surrounding UI; maintain a redaction log for any crop/blur.
• Video → hash the file; log duration; keep original audio; avoid jump-cut edits unless logged.
• Webpage → Wayback + archive.today; save raw HTML; capture key screens after login if allowed; include rendered + source.
• Text/PDF → hash the file; include author, version, export settings; if OCR, attach the OCR log.

Buttons & behaviors

• Inspect Chain → builds the checklist, “What we have,” score, and share-safe summary.
• Export MD / JSON → drop into a public/private receipt page.
• Archive @ Wayback / archive.today → opens both with your URL.
• Share link → preloads fields for collaborators.
• Download HTML (if enabled in your build) → one-click, branded receipt to host.

Presets (grab and go)

• Public proof lite
  • Intended use: Public post · Type: Image
  • Claims: “Screenshot is unedited except crop”, “From internal tool”
  • Tick: Source archived; Redaction log; Hash recorded
  • Goal: Score ≥70; share receipt link with the post
• Press kit evidence
  • Intended use: Press / legal · Type: Video
  • Claims: “Demo shows real system behavior”
  • Tick: Signed at capture; Hash; Context; Timestamp & operator; Public receipt link
  • Goal: Score ≥90; include full-res video + raw logs
• Internal sanity check
  • Intended use: Internal decision · Type: Webpage
  • Claims: “Policy changed on July 1”
  • Tick: Archive both versions; Context captured
  • Goal: Score ≥60; receipt page in wiki

Mental model

Pixels lie; chains don’t. Make lying expensive and honesty cheap: sign at capture, hash everything, archive the source, and publish a single receipt link you can reference forever.