#6464th of 73 in Skills, plugins and rules

agent-skills-for-context-engineering

Collection of 18 Agent Skills on context and harness engineering for coding agents

Stars
18.1k
License
MIT
Last commit
Oct 2026
Last release
May 2026
Language
Python
Stars, last 29 weeks+4,049
25 Mar 2026: 14,087 stars1 Apr 2026: 14,327 stars8 Apr 2026: 14,629 stars15 Apr 2026: 14,866 stars22 Apr 2026: 15,028 stars29 Apr 2026: 15,148 stars6 May 2026: 15,291 stars13 May 2026: 15,447 stars20 May 2026: 15,644 stars27 May 2026: 15,907 stars3 Jun 2026: 16,148 stars10 Jun 2026: 16,311 stars17 Jun 2026: 16,454 stars24 Jun 2026: 16,582 stars1 Jul 2026: 16,697 stars8 Jul 2026: 16,881 stars15 Jul 2026: 17,138 stars22 Jul 2026: 17,269 stars29 Jul 2026: 17,394 stars5 Aug 2026: 17,498 stars12 Aug 2026: 17,614 stars19 Aug 2026: 17,693 stars26 Aug 2026: 17,772 stars2 Sept 2026: 17,847 stars9 Sept 2026: 17,893 stars16 Sept 2026: 17,942 stars23 Sept 2026: 17,980 stars30 Sept 2026: 18,024 stars7 Oct 2026: 18,078 stars14 Oct 2026: 18,136 stars

14,087 on 25 Mar 202618,136 on 14 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

A set of SKILL.md directories covering context fundamentals, compression, memory, tool design, multi-agent patterns, evaluation and harness design. It installs as a Claude Code plugin marketplace entry, an Open Plugins manifest for Cursor and Codex, or by copying single skill folders into .claude/skills, .cursor/skills or .codex/skills. Six worked examples and a router benchmark are included.

Who it is for: Engineers building or tuning coding agents and agent harnesses

Strengths

  • Installs via one /plugin command or by copying individual skill directories
  • Works across Claude Code, Cursor, Codex and other Open Plugins hosts
  • Skill router benchmarked: 600 calls per sweep across four models, with published results
  • Includes six worked examples, such as an LLM-as-judge TypeScript implementation with 19 tests

Weaknesses

  • Skills are guidance text; scripts are Python pseudocode, not a runnable library
  • Router top-1 accuracy ranges 0.84 to 0.92 across tested models
  • Skill activation depends on the host agent's skill discovery
  • Some skills are labeled NEW and their maturity is unknown

What it needs

  • no GPU
  • Needs Claude Code, Cursor, Codex or another Agent Skills host
  • Models: Model-agnostic; router benchmarked on gemini-3.1-pro, composer-2, gpt-5.5, claude-opus-4-7

Also in Skills, plugins and rules

See all 73
Also in Skills, plugins and rules
RankProjectScore
1skillsAgent skills for planning, TDD, debugging and code review291.3k starsMIT94 out of 100
2eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses279.5k starsMIT94 out of 100
3cavemanCuts token use in coding agents with terse replies and compressed inputs Live demo ↗ (opens in a new tab)111.9k starsApache-2.093 out of 100
#4ponytailSkill prompt that makes coding agents write less code and add tests163.6k starsMIT92 out of 100
#5ui-ux-pro-max-skillDesign-system skill that gives coding agents searchable UI styles, palettes and fonts135.9k starsMIT91 out of 100