Fix the license screen for AI-written pull requests

You maintain an Apache-2.0 command line tool, and a contributor opened a pull request with a module an AI agent generated. A model can reproduce licensed code with no attribution, and GPL code inside your project would break the license your enterprise users depend on. So every pull request now goes through a license scan before it merges.

An assistant wrote screen_ai_contribution(matched_licenses, match_scores, has_attestation, threshold), which turns the scan report into a verdict. It is in the editor. It has a bug. Find it and fix it.

Input and output:

• matched_licenses and match_scores: each license the scanner matched, and how strong the match is as a percent, at the same index.
• has_attestation: the contributor signed that they have the right to contribute the code under your license.
• threshold: a match counts only when its score is at least this.
• Return one verdict: reject_copyleft, manual_review, needs_attribution, needs_attestation or accept.

Rules the tests check:

• If the two lists differ in length, return manual_review.
• Ignore matches that score below threshold.
• Uppercase the ASCII letters of each license id, then sort it into a family:
  • copyleft: it starts with AGPL, LGPL, GPL or SSPL
  • permissive: it starts with MIT, BSD, ISC, APACHE or ZLIB
  • unknown: anything else
• Return the first of these that applies:
  1. Any counted match is copyleft: reject_copyleft.
  2. Any counted match is unknown: manual_review.
  3. Any counted match is permissive: needs_attribution.
  4. There is no attestation: needs_attestation.
  5. Otherwise: accept.