Report window: 2026-03-31 → 2026-04-07 | Generated: 2026-04-07T03:28Z
Summary
Overall status: Flat
The labelling pipeline is running reliably — all 8 Label Discussions workflow runs in the last 7 days succeeded, and the correction backlog is clear (0 open signals). However, 6 of 8 daily summary issues did not parse cleanly enough to extract reviewed/changed counts, leaving most throughput data unavailable. There is no previous-window data to compare against, so a directional trend cannot be established with confidence.
Key Metrics
| Metric |
Value |
Note |
| Discussions reviewed (last 7 days) |
≥ 50 |
Partial — only 2 of 8 summaries parsed |
| Label changes applied (last 7 days) |
≥ 6 |
Partial — only 1 of 8 summaries parsed |
| Change rate (last 7 days) |
~60% (6/10) |
From Mar 31 run only; treat as indicative |
| vs. previous 7-day window |
No comparison |
Zero summaries in prior window |
| Correction-collector runs (last 7 days) |
3 |
All succeeded; clustered on 2026-03-31 |
| Open correction signals |
0 |
Backlog clear |
Correction Pressure
Correction pressure is low and resolved. Two correction signals were created on 2026-03-31 and both closed by 2026-04-03 (~3 days to resolution):
#29 — Discussion #98 ("How do I debug GitHub Actions matrix builds failing only on arm64?") → latest label: Actions#27 — Discussion #118 ("Zero support from Github") → latest label: bug
Both signals were grouped under a single parent intake issue (#26 – Batch 01, now closed). This indicates the Correction Collector successfully batched and processed its first intake batch. Raw pressure is minimal and concentrated in that single batch rather than spread across many parents.
No category metadata was attached to either signal, so cluster analysis by category is not currently possible. The two corrected labels (Actions and bug) are distinct — no single label is yet showing a pattern of repeated mis-assignment.
Open Instruction Debt
The correction backlog is clear — 0 open signals, 0 open parent intake issues. The single closed batch (Batch 01) resolved promptly. No instruction debt appears stale at this time.
⚠️ This may reflect that the system is early-stage rather than that labelling instructions are highly accurate. Only 2 total correction signals have been recorded in the last 30 days, which is a low baseline for a healthy signal volume.
The Labelling Correction Feedback workflow ran 15 times in the last 7 days but 14 of 15 runs were skipped (only 1 succeeded on 2026-03-31). This pattern suggests the workflow's trigger conditions are rarely met in practice. This warrants review — either the skip logic is overly broad, or there is genuinely little feedback to process.
Recommendations
-
Fix daily summary issue parsing. 6 of 8 summary issues in the last 7 days yielded no reviewed or changed counts. Without consistent parsing, labelling throughput cannot be tracked. Review the summary issue body format and update the parser to handle all observed title/body patterns (e.g., "Discussion Label Run", "Discussion Label Updates").
-
Investigate Labelling Correction Feedback skip rate. The workflow was triggered 15 times but skipped 14/15. Review the job conditions to confirm they reflect the intended trigger logic; if the workflow is responding to all issue events but only needs to act on specific issue types, tighten the filter to reduce noise.
-
Verify the Correction Collector is still configured to run. All 3 collector runs were on 2026-03-31 — none since. If the collector is supposed to run on a schedule or in response to new correction signals, confirm it is still enabled and correctly targeted.
-
Expand category tagging in correction signals. Neither signal in the current dataset has a category field populated. Adding category metadata (e.g., labelling rule category, discussion type) would allow cluster analysis and earlier identification of systemic mis-labelling patterns. Consider updating .github/instructions/community-discussion-labeling.md to cross-reference signal categories if Actions or bug corrections recur.
Recent daily summary issues (last 7 days)
| Issue |
Date |
Reviewed |
Changed |
State |
#40 — Discussion Label Run — 2026-04-06 |
2026-04-06 |
not parsed |
not parsed |
open |
#38 — Discussion Label Updates – 2026-04-05 |
2026-04-05 |
not parsed |
not parsed |
closed |
#36 — Daily Auto-labelling Summary – 2026-04-04 |
2026-04-04 |
40 |
not parsed |
closed |
#34 — Daily Auto-labelling Summary – 2026-04-03 |
2026-04-03 |
not parsed |
not parsed |
closed |
#33 — 2026-04-03 |
2026-04-03 |
not parsed |
not parsed |
closed |
#31 — 2026-03-31 |
2026-03-31 |
10 |
6 |
closed |
#28 — 2026-03-31 |
2026-03-31 |
not parsed |
not parsed |
closed |
#25 — Daily Auto-labelling Summary – 2026-03-31 |
2026-03-31 |
not parsed |
not parsed |
closed |
Note: Three summaries were created on 2026-03-31 (likely duplicates from early workflow runs). The title format has varied across issues — contributing to parsing failures.
Open correction signal breakdown
No open correction signals at time of report. The full set of 2 signals from the last 30 days:
| Issue |
Created |
Closed |
Latest Label |
#29 — Discussion #98 |
2026-03-31 |
2026-04-03 |
Actions |
#27 — Discussion #118 |
2026-03-31 |
2026-04-03 |
bug |
Parent intake: #26 – Labelling Correction Intake: Batch 01 (closed)
Recent workflow run references
Label Discussions (8 runs, all success):
- run#30 — 2026-04-06 — success
- run#29 — 2026-04-05 — success
- run#27 — 2026-04-03 — success
Labelling Correction Collector (3 runs, all success):
- run#10 — 2026-03-31 — success
- run#9 — 2026-03-31 — success
- run#7 — 2026-03-31 — success
Labelling Correction Feedback (15 runs, 1 success / 14 skipped):
- run#17 — 2026-03-31 — success
- run#27 — 2026-04-06 — skipped
- run#26 — 2026-04-06 — skipped
References
- §30 — Label Discussions run#30 (2026-04-06, success)
- §10 — Labelling Correction Collector run#10 (2026-03-31, success)
- §17 — Labelling Correction Feedback run#17 (2026-03-31, success)
Generated by Labelling Health Report · ● 364.5K · ◷
Report window: 2026-03-31 → 2026-04-07 | Generated: 2026-04-07T03:28Z
Summary
Overall status: Flat
The labelling pipeline is running reliably — all 8
Label Discussionsworkflow runs in the last 7 days succeeded, and the correction backlog is clear (0 open signals). However, 6 of 8 daily summary issues did not parse cleanly enough to extract reviewed/changed counts, leaving most throughput data unavailable. There is no previous-window data to compare against, so a directional trend cannot be established with confidence.Key Metrics
Correction Pressure
Correction pressure is low and resolved. Two correction signals were created on 2026-03-31 and both closed by 2026-04-03 (~3 days to resolution):
#29— Discussion#98("How do I debug GitHub Actions matrix builds failing only on arm64?") → latest label:Actions#27— Discussion#118("Zero support from Github") → latest label:bugBoth signals were grouped under a single parent intake issue (
#26– Batch 01, now closed). This indicates the Correction Collector successfully batched and processed its first intake batch. Raw pressure is minimal and concentrated in that single batch rather than spread across many parents.No category metadata was attached to either signal, so cluster analysis by category is not currently possible. The two corrected labels (
Actionsandbug) are distinct — no single label is yet showing a pattern of repeated mis-assignment.Open Instruction Debt
The correction backlog is clear — 0 open signals, 0 open parent intake issues. The single closed batch (Batch 01) resolved promptly. No instruction debt appears stale at this time.
The
Labelling Correction Feedbackworkflow ran 15 times in the last 7 days but 14 of 15 runs were skipped (only 1 succeeded on 2026-03-31). This pattern suggests the workflow's trigger conditions are rarely met in practice. This warrants review — either the skip logic is overly broad, or there is genuinely little feedback to process.Recommendations
Fix daily summary issue parsing. 6 of 8 summary issues in the last 7 days yielded no
reviewedorchangedcounts. Without consistent parsing, labelling throughput cannot be tracked. Review the summary issue body format and update the parser to handle all observed title/body patterns (e.g., "Discussion Label Run", "Discussion Label Updates").Investigate
Labelling Correction Feedbackskip rate. The workflow was triggered 15 times but skipped 14/15. Review the job conditions to confirm they reflect the intended trigger logic; if the workflow is responding to all issue events but only needs to act on specific issue types, tighten the filter to reduce noise.Verify the Correction Collector is still configured to run. All 3 collector runs were on 2026-03-31 — none since. If the collector is supposed to run on a schedule or in response to new correction signals, confirm it is still enabled and correctly targeted.
Expand category tagging in correction signals. Neither signal in the current dataset has a
categoryfield populated. Adding category metadata (e.g., labelling rule category, discussion type) would allow cluster analysis and earlier identification of systemic mis-labelling patterns. Consider updating.github/instructions/community-discussion-labeling.mdto cross-reference signal categories ifActionsorbugcorrections recur.Recent daily summary issues (last 7 days)
#40— Discussion Label Run — 2026-04-06#38— Discussion Label Updates – 2026-04-05#36— Daily Auto-labelling Summary – 2026-04-04#34— Daily Auto-labelling Summary – 2026-04-03#33— 2026-04-03#31— 2026-03-31#28— 2026-03-31#25— Daily Auto-labelling Summary – 2026-03-31Note: Three summaries were created on 2026-03-31 (likely duplicates from early workflow runs). The title format has varied across issues — contributing to parsing failures.
Open correction signal breakdown
No open correction signals at time of report. The full set of 2 signals from the last 30 days:
#29— Discussion#98Actions#27— Discussion#118bugParent intake:
#26– Labelling Correction Intake: Batch 01 (closed)Recent workflow run references
Label Discussions (8 runs, all success):
Labelling Correction Collector (3 runs, all success):
Labelling Correction Feedback (15 runs, 1 success / 14 skipped):
References