RhetorLint · Claim Feedback 0.2

Words can come back.

Carry claims, not people. Expose a supplied mismatch without pretending to read a soul. Make correction easier than denial.

The page request and static files come from Cloudflare. After those files load, this worksheet processes your selected bytes in browser memory. The app has no claim-upload, fetch, storage, message, signing, dataset, or model endpoint.

One local turn

Review one supplied record

Maximum 65,536 UTF-8 bytes. Selecting or pasting does not run the review. A browser file choice is a selected byte snapshot, not the CLI's stable-path receipt. Stop discards a pending result; an already-started browser digest or file read may still finish before another run can begin.

Waiting. Nothing runs until you choose Run.

One finite path

  1. Exact claimKeep the words, source, scope, and digest together.
  2. Supplied receiptRecord what a named crawler says it observed without authenticating that identity.
  3. Evidence challengeSeparate literal evidence from the interpretation placed on it.
  4. Reply or correctionAttribute the response and append the changed claim without erasing the first.
  5. StopKARMA stays unsigned. Training stays held for independent rights and privacy review.

Stronger local file boundary

The Node door remains

The command-line adapter opens one regular non-symlink file, checks it before and after reading, prints one packet, and stops. The browser and Node adapters share the same deterministic projection. 

npm run claim-feedback -- INPUT.json

Read the complete local guide Open the source repository

For tools and crawlers

Public contracts, not harvested submissions

These reviewed static bytes are deliberately easy to retrieve. The example is fictional and uses reserved example.org addresses. Crawler access is not proof of reuse permission, dataset selection, training, model update, or correction uptake. The schemas describe the JSON shape; runtime checks add cross-field boundaries.

What this does not decide