[deep-report] Optimize Daily Syntax Error Quality Check: add pre-step to cut 56→35 turns and save ~1.36M tokens/run #28747

Description

The Daily Syntax Error Quality Check workflow is consuming 3.4M tokens per run across 56 turns — 40–60% above the expected domain baseline (30–35 turns). Half those turns are pure data-gathering (Phase 1 workflow discovery), which can be replaced with a deterministic bash pre-step.

Evidence: Run §24991120216 — 56 turns, resource_heavy_for_domain: HIGH, agentic_fraction=0.50.

Three concrete edits to the workflow:

1. Add bash pre-step before agent job: find .github/workflows -name '*.md' -type f ! -name 'daily-*.md' ! -name '*-test.md' > /tmp/gh-aw/agent/candidates.txt — eliminates ~8–10 Phase 1 discovery turns
2. Move Install gh-aw CLI from agent prompt into a workflow steps: pre-step — saves ~3–5 turns
3. Add max-turns: 35 to frontmatter and update the turn target in the prompt to ≤ 30 turns

Expected Impact

~1,360,000 tokens saved per run. With daily cadence: ~9.5M tokens/week reduction.

Suggested Agent

Agentic Optimization Kit / manual PR

Estimated Effort

Quick (< 1 hour)

Data Source

DeepReport Intelligence Briefing — April 27, 2026 (run §25003521428); Weekly Agentic Optimization Report #28742

Generated by DeepReport - Intelligence Gathering Agent · ● 473.9K · ◷

• expires on Apr 29, 2026, 3:35 PM UTC