[deep-report] Optimize token usage in CI Cleaner and Agent Persona Explorer workflows #11325
Description
Description
The daily Copilot token report shows CI Cleaner (13.7M tokens) and Agent Persona Explorer (12.5M tokens) dominating 30-day spend with multi‑million tokens per run. Audit their prompts, context size, and data ingestion paths, then implement token budget limits and caching to reduce repeated processing. Prioritize quick wins like smaller diff windows and trimming redundant context.
Expected Impact
Lower monthly Copilot token costs and reduce run time variance while keeping high success rates.
Suggested Agent
Copilot Session Insights or CI Failure Doctor agent
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport Intelligence Briefing 2026-01-22; Token Consumption Report discussion #11248.
AI generated by DeepReport - Intelligence Gathering Agent