loopbacklisted

# Loopback: work the feedback queue Loopback is the feedback bus across this machine's projects. Real product usage — pinned UI feedback, backend failures with response bodies, UX papercuts, feedback on AI features with their run metadata — lands in one queue. Your job is to close loops: fix what real usage surfaced, and write the outcome back so the reporter sees the pin turn green on their page. The project slug: use the one stated in this repo's AGENTS.md ("Working the Loopback queue" section) or in this file if a "Project slug for this repo" line is present below the frontmatter. If neither exists, ask, or check `loopback_get_stats()` for known projects. ## The loop Work one item at a time, most severe first: 1. `loopback_list_feedback(project=<slug>, status="open")` — see what users actually hit. Check `triaged` too if the open queue is empty. 2. `loopback_claim_feedback(id, agent=<your name>)` — claim as your own agent name: `claude-code`, `codex`, `gemini`, or your CLI's name. If the claim is rejected, another agent holds it — pick a different item. Never force-claim (`force=true`) without first stating why in a `loopback_add_comment`. 3. `loopback_get_feedback(id)` — read ALL captured context before touching code. A pin is an anchor, not a scope: a pin on a form button often carries the backend root cause. Specifically read: - `network` and `extra.failed_responses` — failing calls with up to 2KB of response body (error codes, hints, stac