[Code Quality] Fix incorrect percentage calculation in Copilot Agent Analysis report #50433

Description

The copilot-agent-analysis.md workflow prompt computes an incorrect success-rate percentage. In the 2026-08-04 run (discussion #50367), the report states "Merged: 39 (63%)" and shows "62.9%" in the metrics table for 39/71 merged PRs. The correct value is 39/71 = 54.9%, not 62.9%/63%.

This was flagged by the Daily Regulatory Report #50406, which cross-checks daily/periodic report discussions for internal consistency. It notes the error makes the reported "4th consecutive daily decline" trend look less steep than it actually is (true drop is 87.1% -> 54.9%, not 87.1% -> 62.9%).

Suggested Changes

• Review the percentage/success-rate calculation instructions in .github/workflows/copilot-agent-analysis.md (around lines 234 and 270, where [percentage]% is derived from agent_prs_merged / agent_prs_total)
• Ensure the prompt/logic explicitly specifies merged / total * 100 rounded correctly, and add a worked example so the agent computing this at runtime doesn't round or truncate incorrectly
• After editing, run make recompile to regenerate the lock file

Files Affected

• .github/workflows/copilot-agent-analysis.md
• .github/workflows/copilot-agent-analysis.lock.yml (regenerated)

Success Criteria

• Given merged=39, total=71, the workflow computes and reports 54.9% (not 62.9%/63%)
• No regression in other percentage-based fields in the same report

Source

Extracted from Daily Regulatory Report - 2026-08-04 (#50406), which reviewed Copilot Agent Analysis #50367

Priority

Medium - Self-contained data-accuracy bug affecting trend reporting, not blocking.

🔍 Task mining by Discussion Task Miner - Code Quality Improvement Agent · auto · 85.3 AIC · ⌖ 4.64 AIC · ⊞ 10.3K · ◷

• expires on Aug 5, 2026, 5:19 PM UTC-08:00