kzndotsh/skills

Agent skills for Claude Code and other agents that read SKILL.md. Each folder under skills/ is one skill: a SKILL.md entry point plus optional references/ that the agent loads on demand.
Skills
| Skill | What it does |
|---|---|
| prompt-engineering | Audited knowledge base for writing and debugging LLM prompts: technique picker (few-shot, CoT, self-consistency, ReAct, RAG, chaining, PAL, ToT, Reflexion), agents and context engineering, reasoning-model prompting, prompt-injection defenses, model notes, and a prompt library. |
| ai-engineering-from-scratch | Navigator and trust guide for the AI Engineering from Scratch curriculum (542 lessons, 20 phases): a per-phase lesson index, learning routes, a glossary index, and an audit that flags which lessons are correct, stale, or flawed. |
| arctic-shift | Query archived Reddit posts, comments, threads, subreddits, and users through the free Arctic Shift API when reddit.com blocks automated access: endpoint picker, scoping rules, pagination, and failure modes. |
Install
With the skills CLI (hosted at kzndotsh/skills):
Or by hand, from a clone:
Use with Ruler
Ruler keeps one set of agent instructions and skills in a project's .ruler/ folder and copies them into each agent's native location (.claude/skills/, .agents/skills/, .cursor/skills/, and others). These skills use the plain /SKILL.md layout it expects.
Skills support in Ruler is experimental and on by default. Preview with --dry-run. Commit .ruler/ and treat the generated agent directories as build output.
Other agents
Skills in this layout also work in Cursor, which reads .agents/skills/, .cursor/skills/, and the Claude and Codex compatibility paths (.claude/skills/, .codex/skills/), at project or user level. Its docs require name to match the folder name, which holds here. Cursor cannot import a single skill from a repo. To share with a Cursor team, package skills as a plugin with .cursor-plugin/marketplace.json; that isn't set up here.
Adding a skill
Copy template/ to skills/ / and edit.
Lint
Checks every skill under skills/ against the Agent Skills spec and Claude's authoring limits (name and description rules, reserved words, valid frontmatter YAML, body under 500 lines, contents lists, broken or unlinked references) and the llms.txt files (structure, links into this repo, freshness of the root file). CI runs it on every push. No dependencies.
llms.txt
llmstxt.org-style indexes for agents that can fetch a URL but can't install skills:
- llms.txt: the master index. Lists each skill and links to that skill's own index.
- skills/ /llms.txt: one per skill, inside the skill folder. It links the skill's SKILL.md, every reference file and any scripts by raw URL. For ai-engineering-from-scratch it is one large file (about 140 KB, roughly 35k tokens) with a section per phase linking all 542 upstream lessons at a pinned commit, each with our audit status.