Fix silent conversation-transcript fetch failures in Copilot session data fetcher

[Copilot]

copilot-session-data-fetch.md has produced zero *-conversation.txt transcripts for 71+ consecutive days, blocking turn-by-turn tool use, token counts, loop detection, and prompt-quality analysis in Copilot Session Insights — reporting has silently degraded to run-metadata-only.

Root cause

Copilot coding-agent runs execute under a special dynamic/copilot-swe-agent/copilot workflow path. Their job logs do contain the expected [cca-engine] turn=N ... lines (confirmed on real runs), but the fetch script's gh api calls to list jobs and download job logs discarded all stderr/exit codes (2>/dev/null || true). Once the default GITHUB_TOKEN lost the ability to fetch these logs, every failure went completely unnoticed — the logs/ directory stayed empty with no signal as to why.

Changes

• Elevated token fallback: use secrets.GH_AW_GITHUB_TOKEN || secrets.GITHUB_TOKEN, matching the existing repo convention for operations that need permissions beyond the default token.
• Exit-code-based diagnostics: capture stderr and exit codes for both the job-listing and job-log-download gh api calls; classify each failure (API error, no jobs found, empty log, no matching lines) by actual exit code rather than stderr presence, and log the specific reason per run.
• Summary warning: when zero transcripts/events are retrieved across all runs, emit a breakdown by failure category with remediation guidance (configure GH_AW_GITHUB_TOKEN).
• Docs: updated "Session Log Access" and "Requirements" sections to describe the token fallback chain, the new diagnostics, and the documented root cause.
• Recompiled the two consumers of this shared component: copilot-session-insights.md and copilot-opt.md.

# before: failure was invisible
gh api "repos/$REPO/actions/runs/$run_id/jobs" --jq '.jobs[].id' 2>/dev/null || true

# after: failure is captured, classified, and surfaced
job_ids="$(gh api "repos/$REPO/actions/runs/$run_id/jobs" --jq '.jobs[].id' 2>"$jobs_err")" || jobs_rc=$?
if [ -z "$job_ids" ] && [ "$jobs_rc" -ne 0 ]; then
  echo "Failed to list jobs for run $run_id (exit=$jobs_rc): $(head -c 300 "$jobs_err")" | tee -a "$FETCH_STATUS_LOG"
  API_ERROR_COUNT=$((API_ERROR_COUNT + 1))
fi

• Fixes [deep-report] Root-cause the 71+ day conversation-transcript fetch failure blocking Copilot Session Insights #51113

[Copilot]

Pull request overview

Improves Copilot session transcript retrieval and failure visibility.

Changes:

• Adds elevated-token fallback and exit-code diagnostics.
• Emits failure summaries with remediation guidance.
• Recompiles both consuming workflows.

Show a summary per file

File	Description
.github/workflows/shared/copilot-session-data-fetch.md	Adds authentication fallback, diagnostics, and documentation.
.github/workflows/copilot-session-insights.lock.yml	Regenerates the session-insights workflow.
.github/workflows/copilot-opt.lock.yml	Regenerates the optimization workflow.

Review details

Tip

Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

• Files reviewed: 3/3 changed files
• Comments generated: 2
• Review effort level: Balanced