[rendering-scripts] Extract token counts from new Copilot CLI footer

[github-actions]

Summary

The Copilot CLI 1.0.48 pretty-print log now emits a single-line footer such as

Tokens    ↑ 163.9k • ↓ 567 • 149.2k (cached)

with no preceding Breakdown by AI model: section. parse_copilot_log.cjs already recognised this line and stripped it from the agent text body, but the actual token counts were never extracted — so generateInformationSection rendered with no token usage at all.

What changed

• parse_copilot_log.cjs: parse the Tokens ↑X • ↓Y • Z (cached) line into input_tokens, output_tokens, and cache_read_input_tokens on the result entry. k suffixes are normalised the same way as the existing model-breakdown path.
• parse_copilot_log.test.cjs: two new tests (with-cached, without-cached) and the existing suppression test still passes.

How this was discovered

Daily verifier ran the parser against run §26328056329 which had inspected the most recent agentic run, §26328075922 (Weekly Editors Health Check, copilot engine, CLI version 1.0.48). The agent-stdio log contained the new footer, the parser produced usage: {}, and the Information section dropped tokens.

After the fix the Information section renders:

**Token Usage:**
- Total: 313,667
- Input: 163,900
- Cache Read: 149,200
- Output: 567

Test plan

• npm run test:js -- parse_copilot_log — 35 passed (33 prior + 2 new)
• npm run test:js -- render_template log_parser_shared parse_copilot_log — 222 passed
• Re-ran the parser against the real run-26328075922/agent-stdio.log and confirmed the Information section now shows the expected token totals.

References

• Verifier run: §26328056329
• Source run inspected: §26328075922

Generated by ✅ Daily Rendering Scripts Verifier · ● 32.4M · ◷

• expires on May 26, 2026, 8:40 AM UTC