docs: make tier-down delegation a positive default, not a tie-breaker#2
Conversation
Adds a new top-level bullet to CLAUDE.md §"Subagent Strategy" elevating the cheap-tier-first principle from a "when in doubt" tie-breaker (line 209) to an active obligation: before doing work in the main session, ask whether a Haiku/Sonnet subagent could handle it per the rubric, and if yes, spawn that subagent. The main session's tier is typically the most expensive option, so reserving it for work that genuinely needs it is the single biggest cost lever. Applies recursively to spawned subagents.
|
Caution Review failedPull request was closed or merged during review No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughDocumentation updated to instruct Claude Code to proactively delegate work to the cheapest sufficient subagent tier (Haiku or Sonnet) before performing tasks in the main session. The delegation policy is reframed as a direct obligation rather than a fallback, and applies recursively when subagents spawn further subagents. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Review rate limit: 4/5 reviews remaining, refill in 12 minutes. Comment |
Summary
Adds a new top-level bullet to
CLAUDE.md§"Subagent Strategy" that elevates the cheap-tier-first principle from a "when in doubt" tie-breaker (line 209) to an active obligation: before doing work in the main session, ask whether a Haiku/Sonnet subagent could handle it per the rubric, and if yes, spawn that subagent.The main session's tier is typically the most expensive option available, so reserving it for work that genuinely needs it is the single biggest cost lever. The rule applies recursively to spawned subagents.
This complements PR #1 (which rewrote the rubric examples to push more work onto cheaper tiers) by making the act of delegation itself proactive rather than reactive.
Test plan
CLAUDE.md§"Subagent Strategy" lines 195–212 — does the new bullet read coherently against the rubric immediately below it?git diff --stat origin/mainshows exactly one file changed, one insertion, no deletionsSummary by CodeRabbit