Description
The PR Description Updater is a read-only, repo-maintenance workflow (permissions: contents/pull-requests/issues: read, no write-capable safe outputs) yet runs on a frontier model. The Weekly Workflow Analysis flagged it as not needing a frontier model — claude-haiku-4-5 / gpt-4.1-mini would cut per-run cost with no quality loss for diff-summarization work. The same audit shows 3 skills reviewers + updaters dominate action-minutes.
Fix
Set a cheaper model in the frontmatter of read-only maintenance workflows (start with PR Description Updater), then sweep other read-only repo-maintenance / moderate-profile audit workflows for the same downgrade.
Expected Impact
Lower per-run AI cost across the highest-frequency read-only workflows with no functional change.
Suggested Agent
Agentic Token Optimizer / Copilot SWE Agent.
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport 2026-06-29; Weekly Workflow Analysis #42229.
Generated by 🔬 DeepReport - Intelligence Gathering Agent · 212.1 AIC · ⌖ 11.9 AIC · ⊞ 9.8K · ◷
Description
The PR Description Updater is a read-only, repo-maintenance workflow (
permissions: contents/pull-requests/issues: read, no write-capable safe outputs) yet runs on a frontier model. The Weekly Workflow Analysis flagged it as not needing a frontier model —claude-haiku-4-5/gpt-4.1-miniwould cut per-run cost with no quality loss for diff-summarization work. The same audit shows 3 skills reviewers + updaters dominate action-minutes.Fix
Set a cheaper model in the frontmatter of read-only maintenance workflows (start with PR Description Updater), then sweep other read-only
repo-maintenance/moderate-profileaudit workflows for the same downgrade.Expected Impact
Lower per-run AI cost across the highest-frequency read-only workflows with no functional change.
Suggested Agent
Agentic Token Optimizer / Copilot SWE Agent.
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport 2026-06-29; Weekly Workflow Analysis #42229.