Report window: 2026-03-29 → 2026-04-05 (last 7 days) · Generated: 2026-04-05T03:32Z
Summary
Overall status: Mixed — system operational but data coverage is incomplete.
Auto-labelling pipelines are running reliably (6 successful Label Discussions runs in the last 7 days) and the first correction batch was collected and processed within 3 days. However, 4 of 6 daily summary issues failed to yield parsed reviewed/changed counts, and duplicate summary issues were created on two separate days, indicating a format or scheduling inconsistency that limits trend visibility.
Key Metrics
| Metric |
Value |
Notes |
| Discussions reviewed – last 7 days |
≥ 50 |
From 2 of 6 parseable summaries (#31: 10, #36: 40); 4 summaries unparsed |
| Label changes applied – last 7 days |
≥ 6 |
Only summary #31 yielded a changed count; #36 (40 reviewed) did not parse |
| Change rate – last 7 days |
~60% (6/10) |
Based on #31 alone; rate for #36 unknown — treat as lower bound |
| Correction-collector runs – last 7 days |
3 |
All completed successfully |
| Open correction signals |
0 |
Both signals from Batch 01 were closed on 2026-04-03 |
⚠️ Metrics are conservative lower bounds. The actual reviewed/changed counts are likely higher once unparsed summaries are accounted for.
Correction Pressure
Correction pressure is low and concentrated: all 3 signals came from a single intake batch (Batch 01, parent issue #26, created 2026-03-31, closed shortly after). Total of 3 corrections across 2 discussions:
- Discussion
#98 ("How do I debug GitHub Actions matrix builds failing only on arm64?") — category: Other Feature Feedback, Questions, & Ideas — 2 corrections, labels added: Actions, question. The auto-labeller had assigned Code Search and Navigation but trusted signals added Actions and question, suggesting a topic-scope miss.
- Discussion
#118 ("Zero support from Github") — category: Enterprise — 1 correction, label added: bug. Suggests the auto-labeller did not infer bug from a clearly bug-type discussion in the Enterprise category.
There is no single dominant cluster yet, but both signals involve the auto-labeller under-assigning labels. The bug and Actions labels appear to be under-triggered by the current instruction set.
Open correction signal breakdown
No open correction signals as of report time. Both signals were resolved as part of Batch 01.
Open Instruction Debt
The correction backlog is currently at zero — both signals from Batch 01 were closed within ~3 days of creation, and there is 1 parent intake issue (#26), also closed. No new open signals exist.
Oldest open signal age: N/A (none open).
The backlog looks clean but nascent — the system has only been running for ~5 days (first activity: 2026-03-31). The real test will be whether the next correction batch accumulates and is resolved at a similar pace. The two patterns from Batch 01 (under-tagging bug in Enterprise discussions, missing Actions for cross-topic Q&A) are candidates for instruction refinement before the next batch grows.
Recommendations
-
Investigate and fix duplicate daily summary issues. On both 2026-03-31 and 2026-04-03, the Label Discussions workflow created 2–3 summary issues for the same day. This inflates issue count and makes trend tracking unreliable. Review the workflow trigger conditions and add a deduplication guard or idempotency check.
-
Fix summary issue parsing failures. 4 of 6 summary issues (including all duplicates) yielded no reviewed/changed counts. If the report format varies between runs, standardize it so the health data collector can reliably extract metrics.
-
Refine labelling instructions for bug in Enterprise and Actions for cross-topic discussions. Both correction signals point to systematic under-labelling. Update .github/instructions/community-discussion-labeling.md to add explicit rules: (a) apply bug when a discussion in the Enterprise category describes a failure or missing functionality, and (b) apply Actions when the discussion body clearly references GitHub Actions workflows, runners, or YAML syntax regardless of primary category.
-
Monitor for the next correction batch. With 40 discussions reviewed on 2026-04-04 and no correction batch yet collected for that date, a second intake batch may be imminent. Verify the Labelling Correction Collector workflow is scheduled to run and that the Labelling Correction Feedback workflow will process any new signals promptly (10 of 11 recent runs were skipped, which is expected when no signals are open).
Recent daily summary issue breakdown
| Issue |
Title |
Created |
Reviewed |
Changed |
Parsed |
#36 |
Daily Auto-labelling Summary – 2026-04-04 |
2026-04-04T20:54Z |
40 |
— |
Partial |
#34 |
Daily Auto-labelling Summary – 2026-04-03 |
2026-04-03T20:56Z |
— |
— |
❌ |
#33 |
2026-04-03 |
2026-04-03T10:47Z |
— |
— |
❌ |
#31 |
2026-03-31 |
2026-03-31T20:56Z |
10 |
6 |
✅ |
#28 |
2026-03-31 |
2026-03-31T11:02Z |
— |
— |
❌ |
#25 |
Daily Auto-labelling Summary – 2026-03-31 |
2026-03-31T09:07Z |
— |
— |
❌ |
Note: Three summary issues were created on 2026-03-31 and two on 2026-04-03. This is likely a workflow scheduling issue.
Recent workflow run references
| Workflow |
Runs (last 7d) |
Successful |
Skipped/Other |
| Label Discussions |
6 |
6 |
0 |
| Labelling Correction Collector |
3 |
3 |
0 |
| Labelling Correction Feedback |
11 |
1 |
10 skipped |
References
- §28 — Label Discussions, latest successful run (2026-04-04)
- §10 — Labelling Correction Collector, latest run (2026-03-31)
- §17 — Labelling Correction Feedback, only successful run (2026-03-31)
Generated by Labelling Health Report · ● 311.3K · ◷
Report window: 2026-03-29 → 2026-04-05 (last 7 days) · Generated: 2026-04-05T03:32Z
Summary
Overall status: Mixed — system operational but data coverage is incomplete.
Auto-labelling pipelines are running reliably (6 successful
Label Discussionsruns in the last 7 days) and the first correction batch was collected and processed within 3 days. However, 4 of 6 daily summary issues failed to yield parsedreviewed/changedcounts, and duplicate summary issues were created on two separate days, indicating a format or scheduling inconsistency that limits trend visibility.Key Metrics
#31: 10,#36: 40); 4 summaries unparsed#31yielded achangedcount;#36(40 reviewed) did not parse#31alone; rate for#36unknown — treat as lower boundCorrection Pressure
Correction pressure is low and concentrated: all 3 signals came from a single intake batch (Batch 01, parent issue
#26, created 2026-03-31, closed shortly after). Total of 3 corrections across 2 discussions:#98("How do I debug GitHub Actions matrix builds failing only on arm64?") — category: Other Feature Feedback, Questions, & Ideas — 2 corrections, labels added:Actions,question. The auto-labeller had assignedCode Search and Navigationbut trusted signals addedActionsandquestion, suggesting a topic-scope miss.#118("Zero support from Github") — category: Enterprise — 1 correction, label added:bug. Suggests the auto-labeller did not inferbugfrom a clearly bug-type discussion in the Enterprise category.There is no single dominant cluster yet, but both signals involve the auto-labeller under-assigning labels. The
bugandActionslabels appear to be under-triggered by the current instruction set.Open correction signal breakdown
No open correction signals as of report time. Both signals were resolved as part of Batch 01.
#29#98: How do I debug GitHub Actions matrix builds failing only on arm64?Actions#27#118: Zero support from GithubbugOpen Instruction Debt
The correction backlog is currently at zero — both signals from Batch 01 were closed within ~3 days of creation, and there is 1 parent intake issue (
#26), also closed. No new open signals exist.Oldest open signal age: N/A (none open).
The backlog looks clean but nascent — the system has only been running for ~5 days (first activity: 2026-03-31). The real test will be whether the next correction batch accumulates and is resolved at a similar pace. The two patterns from Batch 01 (under-tagging
bugin Enterprise discussions, missingActionsfor cross-topic Q&A) are candidates for instruction refinement before the next batch grows.Recommendations
Investigate and fix duplicate daily summary issues. On both 2026-03-31 and 2026-04-03, the
Label Discussionsworkflow created 2–3 summary issues for the same day. This inflates issue count and makes trend tracking unreliable. Review the workflow trigger conditions and add a deduplication guard or idempotency check.Fix summary issue parsing failures. 4 of 6 summary issues (including all duplicates) yielded no
reviewed/changedcounts. If the report format varies between runs, standardize it so the health data collector can reliably extract metrics.Refine labelling instructions for
bugin Enterprise andActionsfor cross-topic discussions. Both correction signals point to systematic under-labelling. Update.github/instructions/community-discussion-labeling.mdto add explicit rules: (a) applybugwhen a discussion in the Enterprise category describes a failure or missing functionality, and (b) applyActionswhen the discussion body clearly references GitHub Actions workflows, runners, or YAML syntax regardless of primary category.Monitor for the next correction batch. With 40 discussions reviewed on 2026-04-04 and no correction batch yet collected for that date, a second intake batch may be imminent. Verify the
Labelling Correction Collectorworkflow is scheduled to run and that theLabelling Correction Feedbackworkflow will process any new signals promptly (10 of 11 recent runs were skipped, which is expected when no signals are open).Recent daily summary issue breakdown
#36#34#33#31#28#25Note: Three summary issues were created on 2026-03-31 and two on 2026-04-03. This is likely a workflow scheduling issue.
Recent workflow run references
References