メニュー
Pick and re-evaluate the output_config.effort level for a Claude Fable 5 workload. Use when designing a pipeline that mixes routine and hard tasks, when a call feels slow or expensive for what it's doing, or when porting an Opus 4.7/4.8 harness that defaulted to xhigh.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | effort-router |
| description | Pick and re-evaluate the output_config.effort level for a Claude Fable 5 workload. Use when designing a pipeline that mixes routine and hard tasks, when a call feels slow or expensive for what it's doing, or when porting an Opus 4.7/4.8 harness that defaulted to xhigh. |
Effort is the dial that trades intelligence for latency and cost on Fable 5. Two mistakes dominate: carrying over an old model's default effort unexamined, and picking one effort for an entire pipeline instead of per task class.
high is the API default and the right starting point for most analysis, writing, and agentic/coding work.medium (or low when latency is visible to a user) for routine transforms: classification, short edits, formatting.xhigh only for the tasks that actually need it: large migrations, one-way-door changes, multi-day autonomous runs.max almost never — reserve it for problems where xhigh measurably fell short, not as a default upgrade.If you're migrating a harness built for Opus 4.7 or 4.8, don't keep xhigh "because that's what coding needs" — that guidance was for the older model. Re-test at high first; Fable 5 at high frequently matches or beats an older model at xhigh.
A triage step at low/medium that escalates only the cases that need it to high/xhigh beats any single fixed setting across a mixed workload — see job_packet/packet.py in this repo for a working example: prep and execute run cheap, plan_or_audit spends the premium effort only where judgment is dense.
Before shipping a pipeline's effort settings, ask: if I doubled every effort level, would output quality visibly improve? If not, you're overpaying somewhere.
Use before any approved Mythos-style defensive cybersecurity run, or when reviewing a Fable/Mythos security workflow. Requires authorization, bounded targets, private evidence handling, independent validation, and a patch/report owner before the model explores vulnerabilities.
Stop Claude Fable 5 from over-planning, re-deriving settled facts, or listing options it won't pursue. Use in interactive sessions where turns feel slow on simple asks, or agentic sessions at high/xhigh effort where the model gathers more context than the task needs.
Keep Claude Fable 5 running to completion in unattended pipelines -- no stalling on a stated intent without the matching tool call, no mid-run permission questions, no self-imposed session splits from context-budget worry. Use for overnight runs, scheduled jobs, and CI agents where nobody is watching to type "continue".
Bind every progress claim Claude Fable 5 makes during a long or autonomous run to an actual tool result from that session. Use for multi-hour agents, overnight pipelines, CI runs, or any workflow where a status update has previously claimed work that wasn't actually done.
Maintain a file-based lesson memory that Claude Fable 5 reads and updates across sessions. Use for recurring agents (daily jobs, long projects, team assistants) where the same correction keeps being re-made, or when bootstrapping a new agent from past session history.
Keep Claude Fable 5's diffs mapped one-to-one to the request -- no unrequested refactors, speculative abstractions, or defensive error handling for states that can't occur. Use in code review and any coding session where the diff comes back larger than the ask, especially at high or xhigh effort.