From d67f45c6acdb33bc13a4dd815ada9b2738b0ba34 Mon Sep 17 00:00:00 2001 From: Cristian Magherusan-Stanciu Date: Thu, 30 Apr 2026 14:47:16 +0200 Subject: [PATCH] docs: make tier-down delegation a positive default, not a tie-breaker MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 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. --- CLAUDE.md | 1 + 1 file changed, 1 insertion(+) diff --git a/CLAUDE.md b/CLAUDE.md index 993dd61..be385aa 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -202,6 +202,7 @@ Non-trivial work happens in a dedicated git worktree branched off the current br - **When NOT to use subagents**: tight debugging loops where each iteration informs the next, work requiring multiple rounds of your own judgement, interactive refinement with the user. - **Parallel vs background**: multiple independent queries → send them in one message as parallel `Agent` tool calls. Long-running watchers (CI, builds) → `run_in_background: true`, then read output when notified. - **Multi-agent coordination**: see `~/.claude/multi-agent-comms.md` for lock patterns (e.g., `git-push` lock before pushing shared-branch fixes). +- **Default: delegate to the cheapest sufficient tier — actively, not just when in doubt.** Before doing a piece of work in the main session (or spawning a subagent at the same tier as the main session), ask: *can a Haiku or Sonnet subagent handle this per the rubric below?* 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. Treat the rubric as a positive obligation to delegate down, not just a tie-breaker. This applies recursively — when a subagent itself needs to spawn further subagents, it should also default to the cheapest sufficient tier. - **Match model to task complexity via the `Agent` tool's `model` parameter** — pay for capability only when it earns it: - **Haiku** (default for most delegations): file renames, typo fixes, mechanical edits with a clear spec, simple lookups (grep for a symbol, find where X is called), reading a single file to answer a factual question, formatting/style fixes, running a single command (or routine `gh`/`git` operations) and reporting output, implementing a tightly-specified function, writing a new test from a tight spec, code review of a small single-file diff, mechanical API/SDK migration where the mapping is documented, classifying or labelling items against a clear rubric (e.g. backlog triage chunks), summarising a single file or short diff. Cheap, fast, good enough when the answer is mostly mechanical or rubric-driven. - **Sonnet**: focused multi-file changes where coordination across files needs judgement, implementing a function whose spec is mostly clear but has 1–2 design choices, code review of a multi-file diff or a diff with non-trivial logic, refactors with a clear target shape, agentic loops where each step requires non-trivial reasoning (not just rubric-driven labelling). Use when there's some judgement involved but not deep design or hypothesis iteration.