Feature request: First-class persistent findings store for code reviews and audits

Problem

When running code reviews or audits in Claude Code, findings are trapped in the conversation context. Running /clear or starting a new session loses all identified issues, their severity, file locations, and remediation notes.

Current workarounds exist but are manual and fragmented:

• Writing findings to a file mid-review (breaks flow)
• Copying patterns to MEMORY.md after the fact (200-line cap, only for patterns not individual findings)
• Using /obs or similar to export to external tools (requires explicit action per review)

None of these are first-class — they all require the user to remember to persist findings before context is lost.

Proposed Solution

A built-in findings/issues store that:

1. Auto-persists structured findings from code reviews (file path, line range, severity, description, suggested fix)
2. Survives /clear and session boundaries — stored on disk, not just in context
3. Reloads on session start so Claude can reference prior findings ("Last review found 3 unresolved P1 issues")
4. Supports status tracking — open/fixed/wontfix, so findings can be triaged across sessions
5. Exports to GitHub Issues, markdown files, or other formats

Why This Matters

Code reviews are one of the highest-value uses of Claude Code. But the review-to-action pipeline has a gap: findings are ephemeral by default. Users who run large audits (20-30+ findings) risk losing hours of analysis on a single /clear.

A persistent findings store would make Claude Code a credible continuous code review tool, not just a one-shot advisor.

Alternatives Considered

| Approach | Limitation |
|----------|-----------|
| MEMORY.md auto-memory | 200-line cap, unstructured, meant for patterns not individual issues |
| TodoWrite tasks | Session-scoped, lost on /clear |
| Manual file writes | Breaks review flow, user must remember to do it |
| GitHub Issues export | Good for output, but no re-import for status tracking |