Token Optimization Skill Leaderboard

A token optimization toolkit that compresses LLM prompts and agent communications to reduce consumption by up to 65%.

Adjusts AI output for ADHD users by leading with next actions, numbering steps, and suppressing verbose tangents.

A collection of agent skills for context engineering, harness engineering, and the development of production-ready multi-agent systems.

Engineering tools for systematic bug diagnosis and the documentation of root cause analyses.

Authors and structures agent skills according to the agentskills.io spec, including metadata validation and procedural instruction drafting.

Agent orchestration skills for driving browser and desktop GUIs and performing transcript compaction.

A skill library for bulk code refactoring, test fixing, feature planning, codebase auditing, project bootstrapping, and HTML visual documentation generation.

Performs efficiency audits to detect over-engineered code and redundant prose without modifying the source.

A set of AI skills for codebase design, architectural research, and adversarial review to generate precise technical specifications.

Reduces token usage across coding agent sessions through configurable rules for replies, artifacts, tests, code, context management, and delegation, with correctness guardrails.

Dispatches background AI worker agents from checklist-based plans with progress tracking, IPC-based question handling, and config-driven model selection.

A token optimization skill that provides multiple levels of terse, caveman-style or classical Chinese communication to reduce output length.

A routing framework to reduce operational costs by assigning routine agent tasks to low-cost models and complex reasoning to premium tiers.

Audits, distills, and maintains OpenClaw workspace files—AGENTS.md, SOUL.md, TOOLS.md, MEMORY.md, and checklists—for token efficiency and cross-file consistency.