[copilot-session-insights] Daily Copilot Agent Session Analysis — 2026-11-10

[github-actions[bot]]

🤖 Copilot Agent Session Analysis — 2026-11-10

Data-quality warning: no conversation transcripts were available for this run. Turn, token, tool-call, loop, error-recovery and prompt-quality metrics are NOT computed. Figures below come from run-level metadata and orphan-branch data from the GitHub API. Per cache history, this is the sixth consecutive recorded day with this gap.

Executive Summary

• Runs analyzed: 27 coding-agent runs (metadata only; 0 transcripts). The ~50-session target was not met.
• Analysis window: 2026-11-10 00:36Z to 06:40Z (about 6 hours)
• Completion rate: 85.2% overall (23 of 27); 100% of finished runs (23 of 23); 0 failed; 4 still in progress
• Average duration (completed runs, wall clock): 16.5 min; median 14.8 min
• Orphaned branches: 1 of 10 active branches (10.0%); status NORMAL against a 0.0% baseline
• Experimental strategy: none. Transcript-dependent strategies cannot run without transcripts.

Key Metrics

Metric	Value	Trend vs 2026-11-09
Total runs	27	window differs
Successful	23 (85.2%)	↑ (50.0%)
Failed	0	→ (4)
In progress	4	n/a
Avg / median duration	16.5 / 14.8 min	↓ (19.6 / 17.2 min)
Loop detection	not computed	n/a
Context issues	not computed	n/a
Tool failures	not computed	n/a

Run-Level Breakdown by Task Type

Task type (run name)	Runs	Succeeded	Failed	In progress
Addressing comment on PR (7 PRs)	17	17	0	0
Running Copilot cloud agent	10	6	0	4

Trend Charts (run-level, recorded days)

Caveats: the charts use 13 recorded days from cache history (2026-10-29 to 2026-11-10). Earlier rows lack a success/failure split and are excluded. Window sizes vary (n = 10 to 103), so day-to-day comparisons are approximate. The loop overlay is omitted because there are no transcripts.

Success Factors ✅ and Failure Signals ⚠️

Pattern analysis from transcripts is not possible today. Run-level observations only:

1. Comment-addressing runs: 17 of 17 finished runs succeeded across 7 PRs. Their relation to prompt quality cannot be assessed.
2. Failures: none recorded in the window. The 4 in-progress cloud-agent runs (created 06:40Z) are not yet classified.

Orphaned Branch Escalation Alerts 🚨

Branches with at least 5 gate runs, no Copilot assignee on the open PR, and at least 1 hour of waiting.

• Open PRs: 12. Active branches (in-progress runs, last 6 h): 10, carrying 38 in-progress runs.
• Orphaned branches today: 1 of 10 (10.0%)
• Historical baseline: 0.0% (mean over the 12 recorded days 2026-10-29 to 2026-11-09)
• Status: NORMAL (delta +10.0 points; the ELEVATED threshold is more than 20 points)

Branch	PR	Gate count	Wait time	Severity	Recommended action
chore/update-cli-scanner-pins-2026-11-10-e184d07823e9bf87	#76415	12	about 69 min at 06:46Z (created 05:37Z)	🟡 Warning	Assign agent

If PR #76415 is still unassigned at 07:37Z, it reaches the Critical tier (12 gates, more than 2 hours waiting), and the recommended action becomes immediate manual review.

Excluded: copilot/sighthound-fix-security-findings (8 gates) has Copilot assigned, so it is not orphaned. main (6 gates) has no open PR, so it is out of scope. CI waste was not estimated because gate-run durations were not collected.

Data Quality and Diagnostics

• Transcripts: logs/ is empty, the cache session-logs-latest/ directory is empty, and tool-summaries.json reports source=missing with measurements_available=false for all 27 runs.
• Sample size: 27 runs were listed, not the ~50 targeted. The window covers about 6 hours, not 30 days.
• Series break: cache rows from 2026-11-08 onward cover different windows (n=103 on 11-08, n=10 on 11-09, n=27 today). Repo-memory records a method change on 2026-11-08 to coding-agent runs only, which is not comparable with the earlier CI-gate-mix completion figures.
• Duration jump: average duration rose from 1.7 to 5.7 min (Oct 29 to Nov 7) to 16 to 24 min (Nov 8 to 10). The method change above is the likely explanation. Transcripts are needed to confirm whether behaviour changed.

Actionable Recommendations

For System Improvements

1. Restore transcript ingestion: the session-data fetch has produced an empty logs/ directory and source=missing tool summaries for six days. This blocks all behavioural analysis.
   • Potential impact: High
2. Gate the full report on transcript availability: when logs/ is empty, publish a short data-quality note instead of the full behavioural template, so readers do not take absent metrics as zero.
   • Potential impact: Medium
3. State the analysis window in every chart and summary: the target was 50 sessions over 30 days, but this run covers about 6 hours.
   • Potential impact: Low

For Orphaned Branches

4. Assign an agent or review PR [ca] chore: update eligible CLI and scanner pins #76415 manually (12 gates, no assignee). It becomes Critical after 07:37Z.

For Users Writing Task Descriptions

No guidance can be derived from run metadata alone. Re-run this analysis once transcripts are available.

Statistical Summary

Runs listed:               27
Successful:                23 (85.2% of listed; 100% of finished)
Failed:                     0
In progress:                4
Avg duration (completed):  16.5 min
Median duration:           14.8 min
Range (completed):         5.3 to 27.4 min
Loop detection:            not computed (no transcripts)
Orphaned branches:         1 of 10 (10.0%); baseline 0.0% over 12 days

Next Steps

• Restore transcript retrieval in the session-data fetch step
• Assign an agent or owner to PR [ca] chore: update eligible CLI and scanner pins #76415
• Re-run the behavioural analysis once transcripts are present

---

Analysis generated automatically on 2026-11-10 at 06:46Z
Run ID: 42770135540
Workflow: Copilot Session Insights

Generated by 📊 Copilot Session Insights · claude · routed: haiku55 xhigh · ~5 AIC · ⌖ ~2 AIC · ⊞ 15.4K · ◷

• expires on Nov 10, 2026, 10:50 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-11-11T06:50:07.847Z.

Closed by Workflow