Description
find /tmp/gh-aw/agent/session-data/logs -name "*-conversation.txt" has returned 0 files for at least the last ~71 consecutive days (per discussion #51035, 2026-08-07 Copilot Agent Session Analysis), despite the log directory existing. This is distinct from the all-or-nothing artifact-check bug already fixed in #50273 (closed) — that fix addressed the run failing outright when a sampled log was missing, but the underlying fetch gap (zero transcripts ever retrieved) remains unresolved. As a result, turn-by-turn tool use, token counts, loop detection, and prompt-quality analysis have been impossible for over two months; all reporting falls back to workflow-run metadata only.
Expected Impact
Restore actual conversation transcript retrieval so Copilot Agent Session Analysis can detect looping sessions, context issues, and prompt-quality regressions instead of relying solely on run metadata (completion/duration).
Suggested Agent
New Agent (session-data pre-fetch step maintainer)
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport analysis 2026-08-07, discussion #51035 (Daily Copilot Agent Session Analysis).
Generated by 🔬 Deep Report · agent · 156.7 AIC · ⌖ 37.4 AIC · ⊞ 11.5K · ◷
Description
find /tmp/gh-aw/agent/session-data/logs -name "*-conversation.txt"has returned 0 files for at least the last ~71 consecutive days (per discussion #51035, 2026-08-07 Copilot Agent Session Analysis), despite the log directory existing. This is distinct from the all-or-nothing artifact-check bug already fixed in #50273 (closed) — that fix addressed the run failing outright when a sampled log was missing, but the underlying fetch gap (zero transcripts ever retrieved) remains unresolved. As a result, turn-by-turn tool use, token counts, loop detection, and prompt-quality analysis have been impossible for over two months; all reporting falls back to workflow-run metadata only.Expected Impact
Restore actual conversation transcript retrieval so Copilot Agent Session Analysis can detect looping sessions, context issues, and prompt-quality regressions instead of relying solely on run metadata (completion/duration).
Suggested Agent
New Agent (session-data pre-fetch step maintainer)
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport analysis 2026-08-07, discussion #51035 (Daily Copilot Agent Session Analysis).