claude

compute-ladder

Routes Claude Code's compute: Claude escalates its own reasoning effort mid-turn when a task is genuinely hard (with a guard against accidental downgrades), delegates large mechanical jobs to a Sonnet low-effort worker, and reports how it routed

1 on GitHubby xiaolaiv0.2.0ISCupdated 1 month agodeveloper-tools
/

Plugin compute-ladder from marketplace xiaolai in xiaolai/compute-ladder-for-claude. Commands read from the repo's manifest — 1 month ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (3)

  • escalate-max

    Raises effort to max for this turn. Invoke it FIRST when a mistake would be costly or a problem already resisted an attempt - security-sensitive code (authentication, authorization, cryptography, secrets, payments); destructive operations on production data; a production incident; a second failure;

  • escalate-xhigh

    Raises effort to xhigh for this turn. Invoke it FIRST, before reading files, for genuinely hard work - a costly-to-reverse design, architecture, data-model or API decision; debugging a non-obvious cause (concurrency, caching, timing); a multi-file refactor or migration with invariants to preserve; a

  • report

    Reports how compute-ladder has been escalating effort - escalation rate, output tokens per effort tier, escalations that did not take effect, and turns that look under- or over-escalated. Use when the user asks how compute-ladder's routing is performing or whether its routing is working.

Subagents (1)

  • low-worker

    Runs well-specified mechanical work on Sonnet at low effort - bulk edits that follow an explicit pattern, renames across many files, repository-wide searches, format conversions, running commands and reporting their output. Worth delegating when the job is large (many files or long outputs) and need

Hooks (5)

  • SessionStart

    cat "${CLAUDE_PLUGIN_ROOT}/hooks/policy.md"

  • UserPromptSubmit

    node "${CLAUDE_PLUGIN_ROOT}/scripts/prompt-submit.mjs"

  • PreToolUse

    node "${CLAUDE_PLUGIN_ROOT}/scripts/guard.mjs"

  • Stop

    node "${CLAUDE_PLUGIN_ROOT}/scripts/stop.mjs"

  • SubagentStop

    node "${CLAUDE_PLUGIN_ROOT}/scripts/subagent-stop.mjs"

View source on GitHub

Similar plugins

mattpocock-skills

claude

Matt Pocock's agent skills for real engineering (grilling, spec/ticket flows, TDD, code review, domain modelling and more).

39 skills

mattpocock/skills 323,108

academic-research-skills

claude

5 skills + 35 modes + Material Passport pipeline. Includes v3.6.7 cross-model audit gate and v3.6.8 generator-evaluator contract.

5 skills · 16 commands · 4 agents · hooks

Imbad0202/academic-research-skills 56,698

slidev

claude

Claude Code skill for creating and presenting Slidev slidedecks

1 skill

slidevjs/slidev 49,452

llm-finetuning

claudecodex

Eval-gated LLM fine-tuning lifecycle (LoRA/QLoRA, DPO, GRPO/RLVR, vision SFT, quantized export). Use when fine-tuning an open-weights model end-to-end: building the eval harness first, selecting method and model, preparing data, training with Unsloth or TRL, gating checkpoints, and exporting GGUF/FP8.

10 skills · 2 commands · 3 agents

wshobson/agents 42,461

diffusers

claudecodex

Conventions and task skills for contributing to diffusers

4 skills

huggingface/diffusers 34,915

mlflow-tracing

claude

Observability plugin for Claude Code with MLflow tracing

mlflow/mlflow 28,788

Maintain compute-ladder?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex pluginslisted