resolve-flag

Review copyright flags and present recommendations for the user to confirm.

If the user provided context: $ARGUMENTS

1. Gather flags

If the user provided a specific track ID/URL or flag notification, extract the track ID and query just that scan.

Otherwise, fetch all pending flags from the moderation service:

source .env && curl -s "https://moderation.plyr.fm/admin/flags?filter=pending" \
  -H "X-Moderation-Key: ${MODERATION_AUTH_TOKEN}"

Also check the prod database (cold-butterfly-11920742) for recent scans that may not have labels yet:

SELECT cs.id, cs.track_id, cs.is_flagged, cs.highest_score, cs.matches, cs.scanned_at,
       t.title, a.handle, a.display_name
FROM copyright_scans cs
JOIN tracks t ON cs.track_id = t.id
JOIN artists a ON t.artist_id = a.id
WHERE cs.is_flagged = true
ORDER BY cs.scanned_at DESC

2. Analyze each flag

For each flagged scan, classify it using these known patterns:

Self-match (very common)

The primary match artist is the same person who uploaded the track. Compare match artist names against the uploader's handle/display name — variations count (e.g. handle knock2one.bsky.social matching artist knock2one). If the majority of matches are self-matches with score 0, this is a clear false positive.

Recommendation: resolve as ORIGINAL_ARTIST

(add more patterns here as they're encountered)

3. Present recommendations

Use AskUserQuestion to present findings. For each flag, show:

• track title and artist
• what the matches look like (brief summary, not the full list)
• your classification and confidence
• recommended action

Let the user confirm, modify, or skip each recommendation before taking any action.

4. Execute confirmed actions

For flags the user confirms should be resolved:

If a label exists in the moderation service:

source .env && curl -s -X POST "https://moderation.plyr.fm/admin/resolve" \
  -H "X-Moderation-Key: ${MODERATION_AUTH_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"uri":"<ATPROTO_URI>","val":"copyright-violation","reason":"<REASON>","notes":"<notes>"}'

Valid reasons: ORIGINAL_ARTIST, LICENSED, FINGERPRINT_NOISE, COVER_VERSION, OTHER, content_deleted

If no label exists (scan is in DB but moderation service has no label), just report that no action is needed — the system already handled it correctly.

If the user wants to escalate, suggest running the moderation loop: uv run scripts/moderation_loop.py --env prod