Summary
Status: Mixed — pipeline is active but correction backlog is stagnant and feedback loop is broken.
The Label Discussions workflow ran successfully every day last week and over the last 30 days. However, no auto-labelling daily summary issues were parsed, so reviewed/changed counts are unavailable. More critically, the Labelling Correction Feedback workflow has been skipping every run for the entire 30-day window, meaning no correction signals are being resolved. The backlog of 333 open correction signals has not moved in 7+ weeks.
Key Metrics
| Metric |
Value |
| Discussions reviewed — last 7 days |
⚠️ Unknown (no parsed daily summary issues) |
| Label changes applied — last 7 days |
⚠️ Unknown (no parsed daily summary issues) |
| Change rate — last 7 days |
⚠️ Cannot compute |
Label Discussions runs — last 7 days |
✅ 7 / 7 succeeded |
Labelling Correction Collector runs — last 7 days |
⚠️ 0 (not found in data) |
Labelling Correction Feedback runs — last 7 days |
🔴 7 / 7 skipped |
| Open correction signals |
🔴 333 |
| New correction signals — last 7 days |
0 |
| New correction signals — last 30 days |
0 |
| Signals closed — last 7 days |
0 |
Note: No daily summary issues were available in the data (auto_labelling_summaries is empty). Reviewed and changed counts cannot be estimated from this report cycle. This is a data gap that should be investigated.
Correction Pressure
The 333 open correction signals are concentrated in three categories:
| Category |
Open Signals |
| Copilot Conversations |
110 (33%) |
| GitHub Education |
71 (21%) |
| Other Feature Feedback, Questions & Ideas |
52 (16%) |
| Apps, API and Webhooks |
21 (6%) |
| Mobile |
14 (4%) |
| All others |
65 (19%) |
The highest-pressure label clusters by category+label pairing:
- Copilot Conversations →
Copilot: 54 signals
- Copilot Conversations →
Copilot in GitHub: 42 signals
- GitHub Education →
GitHub Education: 38 signals
- Other Feature Feedback →
Other Features and Feedback: 20 signals
The unlabeled event type accounts for 136 of 333 open signals (41%), suggesting the auto-labeller is frequently removing labels that trusted correctors want kept, or applying labels that are subsequently removed.
The five highest-correction individual signals (correction_count ≥ 17) all fall in Other Feature Feedback and GitHub Education categories, pointing to chronic disagreements between the auto-labeller and trusted correctors in those areas.
All 333 signals are grouped under 4 open parent intake issues (all "Batch 02", created 2026-04-08). Pressure is not spread across many batches — it is entirely concentrated in this single unresolved batch.
Open correction signal detail by category
| Category |
Count |
| Copilot Conversations |
110 |
| GitHub Education |
71 |
| Other Feature Feedback, Questions, & Ideas |
52 |
| Apps, API and Webhooks |
21 |
| Mobile |
14 |
| Discussions |
13 |
| Actions |
10 |
| Repositories |
7 |
| Pull Requests |
6 |
| Discover: GitHub Best Practices |
4 |
| Projects and Issues |
3 |
| Codespaces |
3 |
| Packages |
3 |
| GitHub Learn |
3 |
| Accessibility |
3 |
| Code Security |
3 |
| A Welcome to GitHub |
2 |
| npm |
2 |
| New to GitHub |
2 |
| Enterprise |
1 |
Top 5 signals by correction count:
| Signal |
Category |
Current Label |
Corrections |
#114 |
Other Feature Feedback... |
Code Search and Navigation |
19 |
#69 |
GitHub Education |
question |
17 |
#63 |
Other Feature Feedback... |
Profile |
17 |
#55 |
GitHub Education |
GitHub Education Benefits |
17 |
#51 |
Other Feature Feedback... |
Code Search and Navigation |
17 |
Open Instruction Debt
The correction backlog is growing in age, not shrinking. No signals were created or closed in the last 30 days, and the Labelling Correction Feedback workflow has been skipping every run for the full 30-day observation window. This means no automated resolution is occurring.
- Oldest open signal:
#47 — 48 days old
- Open parent intake issues: 4 (all "Batch 02", all 48 days old)
- Closed parent intake issues: 1 ("Batch 01")
- Signal closure rate: 2 closed out of 335 total (0.6%)
The backlog appears stale and unactionable in its current state. The feedback loop that would normally resolve signals is entirely inactive. Without intervention, the backlog will continue aging with no resolution.
Parent intake issue status
| Issue |
Title |
State |
Age |
#50 |
[Labelling Correction Intake] Batch 02 |
🔴 open |
48 days |
#49 |
[Labelling Correction Intake] Batch 02 |
🔴 open |
48 days |
#46 |
[Labelling Correction Intake] Batch 02 |
🔴 open |
48 days |
#45 |
[Labelling Correction Intake] Batch 02 |
🔴 open |
48 days |
#26 |
[Labelling Correction Intake] Batch 01 |
✅ closed |
56 days |
Recommendations
-
Diagnose why Labelling Correction Feedback always skips. This workflow has returned skipped on every one of its 28 runs over the last 30 days. Until this is fixed, no correction signals can be resolved automatically. Inspect the workflow's trigger conditions and any gate checks that may be causing the skip.
-
Investigate the missing Labelling Correction Collector runs. No collector runs appear in the 30-day data. If this workflow is not running, no new signals will be generated going forward. Verify the workflow schedule and trigger configuration.
-
Triage Copilot Conversations labelling rules in .github/instructions/community-discussion-labeling.md. With 110 open signals — 54 resolving to Copilot and 42 to Copilot in GitHub — there is likely ambiguity in how the auto-labeller distinguishes these two labels. Clarifying the instruction boundary between them could reduce future correction pressure.
-
Manually review and close the 4 open Batch 02 parent intake issues (#45, #46, #49, #50) after triaging their sub-issues. A 48-day-old batch with 333 unresolved sub-issues is a signal that the intake-to-resolution pipeline needs an explicit triage cadence, not just automated tooling.
References
Recent workflow runs
Label Discussions (last 7 days — all succeeded)
| Run |
Date |
Conclusion |
| §83 |
2026-05-26 |
✅ success |
| §82 |
2026-05-25 |
✅ success |
| §81 |
2026-05-24 |
✅ success |
| §80 |
2026-05-23 |
✅ success |
| §79 |
2026-05-22 |
✅ success |
| §78 |
2026-05-21 |
✅ success |
| §77 |
2026-05-20 |
✅ success |
Labelling Correction Feedback (last 7 days — all skipped)
Runs #415–#421 (2026-05-20 through 2026-05-26): all skipped.
Labelling Correction Collector: No runs found in the 30-day window.
§83 · §82 · §81
Generated by Labelling Health Report · ● 413.1K · ◷
Summary
Status: Mixed — pipeline is active but correction backlog is stagnant and feedback loop is broken.
The
Label Discussionsworkflow ran successfully every day last week and over the last 30 days. However, no auto-labelling daily summary issues were parsed, so reviewed/changed counts are unavailable. More critically, theLabelling Correction Feedbackworkflow has been skipping every run for the entire 30-day window, meaning no correction signals are being resolved. The backlog of 333 open correction signals has not moved in 7+ weeks.Key Metrics
Label Discussionsruns — last 7 daysLabelling Correction Collectorruns — last 7 daysLabelling Correction Feedbackruns — last 7 daysCorrection Pressure
The 333 open correction signals are concentrated in three categories:
The highest-pressure label clusters by category+label pairing:
Copilot: 54 signalsCopilot in GitHub: 42 signalsGitHub Education: 38 signalsOther Features and Feedback: 20 signalsThe
unlabeledevent type accounts for 136 of 333 open signals (41%), suggesting the auto-labeller is frequently removing labels that trusted correctors want kept, or applying labels that are subsequently removed.The five highest-correction individual signals (correction_count ≥ 17) all fall in Other Feature Feedback and GitHub Education categories, pointing to chronic disagreements between the auto-labeller and trusted correctors in those areas.
All 333 signals are grouped under 4 open parent intake issues (all "Batch 02", created 2026-04-08). Pressure is not spread across many batches — it is entirely concentrated in this single unresolved batch.
Open correction signal detail by category
Top 5 signals by correction count:
#114#69#63#55#51Open Instruction Debt
The correction backlog is growing in age, not shrinking. No signals were created or closed in the last 30 days, and the
Labelling Correction Feedbackworkflow has been skipping every run for the full 30-day observation window. This means no automated resolution is occurring.#47— 48 days oldThe backlog appears stale and unactionable in its current state. The feedback loop that would normally resolve signals is entirely inactive. Without intervention, the backlog will continue aging with no resolution.
Parent intake issue status
#50#49#46#45#26Recommendations
Diagnose why
Labelling Correction Feedbackalways skips. This workflow has returnedskippedon every one of its 28 runs over the last 30 days. Until this is fixed, no correction signals can be resolved automatically. Inspect the workflow's trigger conditions and any gate checks that may be causing the skip.Investigate the missing
Labelling Correction Collectorruns. No collector runs appear in the 30-day data. If this workflow is not running, no new signals will be generated going forward. Verify the workflow schedule and trigger configuration.Triage
Copilot Conversationslabelling rules in.github/instructions/community-discussion-labeling.md. With 110 open signals — 54 resolving toCopilotand 42 toCopilot in GitHub— there is likely ambiguity in how the auto-labeller distinguishes these two labels. Clarifying the instruction boundary between them could reduce future correction pressure.Manually review and close the 4 open Batch 02 parent intake issues (
#45,#46,#49,#50) after triaging their sub-issues. A 48-day-old batch with 333 unresolved sub-issues is a signal that the intake-to-resolution pipeline needs an explicit triage cadence, not just automated tooling.References
Recent workflow runs
Label Discussions (last 7 days — all succeeded)
Labelling Correction Feedback (last 7 days — all skipped)
Runs
#415–#421(2026-05-20 through 2026-05-26): allskipped.Labelling Correction Collector: No runs found in the 30-day window.
§83 · §82 · §81