floccus browser-sync queue shadow pilot

[VrtxOmega]

Hi @marcelklehr,

I ran a private read-only shadow pilot against the public floccusaddon/floccus issue queue, focused on browser-sync reports, feature requests, and reports that need more environment/debug evidence.

No writes to floccus. No comments, labels, PRs, or issue state changes.

Brief here: floccus Browser-Sync Queue Shadow Pilot

Short version: the first dry run got your queue wrong. It treated complete feature requests like incomplete bug reports because they did not include logs, repro steps, or expected/actual sections.

That was wrong for floccus. Your feature-request template asks for use case, requested behavior, and alternatives. We fixed PCF against that live queue shape, reran the pilot, and the split moved from:

• 0 review-now / 9 repair / 3 defer

to:

• 4 review-now / 8 repair / 0 defer

This is not an AI detector and it does not claim perfect triage precision. The narrow question is whether this kind of queue split helps separate review-ready feature requests, reproducible sync bugs, platform/support limitations, and reports that still need logs or environment details.

Worth a look?

---

floccus Browser-Sync Queue Shadow Pilot

Read-only dry run against floccusaddon/floccus.

No writes to floccus. No comments. No labels. No issue state changes.

Queue Split

Sample: 12 open issues from the public issue queue, issue-only.

Run	Review now	Ask for repair / clarification	Defer for now
Initial dry run	0	9	3
After feature-request fix	4	8	0

Repository context was checked for all 12 sampled issues in the final run. Ten items were cleared by repository context, two items had context findings, and no context collection errors occurred.

What PCF Got Wrong

The initial dry run treated complete feature requests like incomplete bug reports.

That was wrong for floccus because the feature-request template asks for:

• the feature or use case
• the requested solution
• alternatives considered

It does not require logs, stack traces, repro steps, or expected/actual bug behavior.

The fix: PCF now recognizes feature requests as a distinct issue shape. It evaluates use case and requested behavior first, while still keeping thin feature requests out of the review-now queue.

Review Now

Issue	Why it surfaced
#2265 hide opened menu when clicking outside	Complete feature request with a concrete UX problem, requested behavior, and alternative.
#2260 Open all in default browser	Complete feature request with use case and requested behavior.
#2259 option to set theme	Complete feature request with use case and requested behavior.
#2257 Notion Integration	Complete feature request with current workflow, use case, requested solution, and alternatives.

Ask For Repair / Clarification

Issue	Why it surfaced
#2251 Android app - e.include is not a function	Repository context found similar closed/solved issues; needs confirmation that the existing fixes do not already cover it.
#2266 Up/ down arrow not working	Repository context found a similar closed/solved issue; needs confirmation that it still reproduces on current versions.
#2274 Can't get past repeated failsafe errors	Report has enough signal to keep, but the queue should route it as clarification/support rather than immediate bug review.
#2269 Renaming a bookmark or folder moves it to the bottom of the list	Reproducible bug shape, but PCF still asks for more explicit expected/actual/log evidence before treating it as fully review-ready.
#2267 Sync never finishes on Orion	Browser-specific sync report that likely needs clearer expected/actual behavior and debug evidence.
#2202 On Brave browser (MacOS), sync icon stucks on syncing	Browser-specific sync report that likely needs clearer current reproduction and debug evidence.
#2246 The bookmarks.xbel file may not be completely written to the WebDAV server	WebDAV sync report with useful context, but still routed as repair/clarification before review.
#886 Fuzzed Testing: Problematic seeds	Long-running technical issue; needs current status/evidence before it should consume immediate review time.

Defer For Now

No sampled issues remained in the defer bucket after the feature-request fix. Thin feature requests are still blocked by the evaluator in the benchmark corpus; this sample simply did not leave any sampled item in that bucket after the fix.

Guardrails

• This was a shadow run against public GitHub data.
• It did not use GitHub write permissions on floccus.
• It did not post comments, apply labels, close issues, or mutate repository state.
• It is not an AI-authorship detector.
• It is not claiming perfect precision over maintainer judgment.
• The useful question is narrower: does this queue split save maintainer attention on browser-sync issues and feature requests?

Reproduce

npm run pilot:public:markdown -- --repository floccusaddon/floccus --limit 12 --no-pulls --write public-pilot.md

Original private pilot artifacts:

/tmp/pcf-floccus-pilot.json
sha256 4f7e8f9b7e235c4c80d02b0d2d6e6caf262b4847a14276024aa81828c5f2c72d

/tmp/pcf-floccus-pilot-final.json
sha256 b5c5da8bf4a16ec7776c0c96f77de928fd4e8b1eef6f3caab799aa90a41e6955

/tmp/pcf-floccus-pilot-final.md
sha256 8ce0781d60f6954c036e668d0367810155556839bd744117b64e749d931a97c3

Implementation Proof

The fix was committed and verified in PCF:

• 3624590 Teach issue evaluator feature-request shape
• e49f324 Upgrade verification workflow actions
• Local gates: 90/90 tests, 38/38 benchmark, 8/8 red-test, maintainer demo PASS.
• GitHub Actions: PCF Verification run 26688220592 completed successfully.