[Meta] Zero maintainer engagement on community-reported bugs — stale bot closes issues never reviewed #39980

What's Wrong

This is a process/trust issue, not a code bug.

I have 21 open issues on this repository, filed between February 28 and March 22, 2026. They document systematic model behavior failures with reproducible examples and evidence from 200+ sessions of professional use.

Zero of these 21 issues have received a comment from anyone at Anthropic.

The community has engaged meaningfully — other users validated bugs, shared workarounds, submitted PRs (#32755), and cross-referenced their own experiences. But there is no signal that any Anthropic engineer, PM, or triage person has read any of them.

The Stale Bot Problem

On March 26, 11 of my issues received the stale label simultaneously. These are:

• [Bug] Claude Code reports task completion without actually executing operations #32281 — Reports completion without executing operations
• MCP MySQL tool parser rejects cross-schema dot notation (schema.table) #32288 — MCP MySQL parser rejects cross-schema dot notation
• [Bug] Claude generates incorrect code/SQL and reports it as complete without validation #32289 — Generates incorrect code/SQL, reports complete
• [Bug] Claude reads files but ignores actionable instructions contained in them #32290 — Reads files but ignores actionable instructions
• [Bug] Verification queries are tautological — cannot return failure results #32291 — Verification queries are tautological
• [Bug] Multi-tab coordination rules in CLAUDE.md are not enforced — tabs duplicate work silently #32292 — Multi-tab coordination rules not enforced
• [Bug] Multi-step tasks lack per-step verification — errors compound silently #32293 — Multi-step tasks lack per-step verification
• [Bug] Claude defaults to assertion from memory instead of verification from tools #32294 — Defaults to assertion from memory instead of verification
• [Bug] Claude silently skips documented verification steps instead of asking the user #32295 — Silently skips documented verification steps
• [Feature] Completion summaries should distinguish verified facts from inferred claims #32296 — Completion summaries should distinguish verified vs inferred
• [Bug] Claude ignores non-fatal tool output — warnings and stderr treated as success #32657 — Ignores non-fatal tool output (warnings/stderr)

All 11 are children of meta issue #32650, which documents a systematic taxonomy of completion-integrity failures across 6 phases. These bugs are not stale — they reproduce daily on Opus 4.6. They are being auto-closed because nobody at Anthropic ever triaged them.

The Impact

I pay for Max plan ($200/mo). I've invested significant time writing detailed, reproducible bug reports with evidence. The implicit contract is: I report bugs with evidence, you at least acknowledge them. A stale bot closing unreviewed issues breaks that contract.

This is not unique to me — see #30407 and #19267 for the same complaint from other users.

What Would Fix This

1. Triage labels — even a triaged or acknowledged label would signal someone read it
2. Stale bot exemptions for issues with community engagement (comments, reactions)
3. Periodic "state of bugs" updates for high-signal meta issues like [Meta] Systematic completion-integrity failures — 16 issues across 6 phases from 100+ sessions of evidence #32650
4. Don't auto-close issues that were never responded to — closing implies resolution; silence followed by closure implies indifference

Environment

• Claude Code v2.1.85, Max plan
• 21 open issues, 200+ sessions of evidence
• 0 maintainer responses across all issues