Investigate zero feedback submissions and declining interactions

[github-actions]

Problem

Feedback submissions are zero in the last 30 days and overall interactions declined; the product lacks user feedback and may have a broken or undiscoverable feedback flow.

Evidence

• Feedback Submissions (30d) = 0 (product data report).
• Total Interactions (7d) dropped from 303 → 231 (≈ -24%).
• Chat activity: no recent messages (7d) and no unique chatters (report).
• my command: 3 uses; myAdded callback: 0 — possible discoverability or callback bug for user-owned-items views.

Expected Behavior

• Feedback flow should accept at least occasional submissions; telemetry should show non-zero submissions when users want to report issues.
• The product team should be able to rely on feedback channel for qualitative signals.

Suggested Investigation Steps (for Project Agent)

1. Verify feedback submission endpoint / webhook and any recent changes to env vars (BOT_AUTH_TOKEN, feedback endpoints, etc.).
2. Check bot logs around feedback endpoints and any errors in past 30 days.
3. Confirm the in-chat UI for feedback (commands, keyboard entries, menus) is present and visible in current bot version.
4. Verify my / myAdded callback flow for discoverability and callback handler errors.

---

Filed by product agent based on telemetry (no user-feedback issues present).

[Szer]

Closing as won't fix — per the repo owner.

The premise is sound but the conclusion isn't actionable: /feedback is genuinely not used much, and that is fine. People prefer to vent in the community chat rather than invoke a command, which is normal behaviour and not a discoverability bug to be engineered around.

This has already been acted on structurally rather than as a one-off fix — .github/prompts/product.md was rewritten to make community-chat mining the primary product signal, with /feedback triage demoted to a secondary path that degrades to a no-op when empty. Evidence for that call: 4 feedback rows all-time, none since 2026-05-10, against a live and chatty community.

Worth noting the strongest product finding in this repo's history came from the chat-reading path, not the command: #266 ("mark inactive" coupon action) quoted real community messages verbatim and shipped the same day as PR #267. The signal was never missing — it just wasn't arriving through the command.

The interactions drop noted here (303 → 231, −24%) is now trended continuously by the monitor role against coupon's own baseline, so a real decline resurfaces with context instead of as a one-off observation.