{"id":51115238,"url":"https://github.com/lroolle/skills","last_synced_at":"2026-06-24T21:01:35.630Z","repository":{"id":364033633,"uuid":"1264796098","full_name":"lroolle/skills","owner":"lroolle","description":"Agent skills with opinions: design, motion, artifacts, decisions, and the craft of making skills themselves.","archived":false,"fork":false,"pushed_at":"2026-06-11T10:05:36.000Z","size":106,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-11T12:06:53.780Z","etag":null,"topics":["agent-skills","ai-agents","claude","claude-code","codex","skills"],"latest_commit_sha":null,"homepage":null,"language":"HTML","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lroolle.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-10T07:29:23.000Z","updated_at":"2026-06-11T10:05:43.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/lroolle/skills","commit_stats":null,"previous_names":["lroolle/skills"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/lroolle/skills","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lroolle%2Fskills","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lroolle%2Fskills/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lroolle%2Fskills/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lroolle%2Fskills/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lroolle","download_url":"https://codeload.github.com/lroolle/skills/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lroolle%2Fskills/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34749211,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-24T02:00:07.484Z","response_time":106,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-skills","ai-agents","claude","claude-code","codex","skills"],"created_at":"2026-06-24T21:01:34.685Z","updated_at":"2026-06-24T21:01:35.624Z","avatar_url":"https://github.com/lroolle.png","language":"HTML","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Skills\n\n\u003e Agent skills with opinions: design, motion, artifacts, decisions,\n\u003e and the craft of making skills themselves.\n\n[![skills.sh](https://skills.sh/b/lroolle/skills)](https://skills.sh/lroolle/skills)\n[![license](https://img.shields.io/github/license/lroolle/skills)](LICENSE)\n[![spec](https://img.shields.io/badge/spec-agentskills.io-blue)](https://agentskills.io/specification)\n\nSmall, composable skills that solve real problems I hit daily with Claude Code\nand Codex. Not a framework. Not a process. Just focused tools you can drop into\nany project and hack on.\n\n## Quickstart\n\n```bash\nnpx skills@latest add lroolle/skills\n```\n\nPick the skills you want and which agents to install them on.\n\n\u003cdetails\u003e\u003csummary\u003eManual install\u003c/summary\u003e\n\n```bash\ngit clone https://github.com/lroolle/skills.git\ncd skills\n./scripts/link-skills.sh\n```\n\nOr copy individual skills:\n\n```bash\ncp -r skills/kiln ~/.claude/skills/kiln\n```\n\n\u003c/details\u003e\n\n## Why these skills exist\n\n### The AI-generated look\n\nMost AI-built interfaces are instantly recognizable: Inter font, purple\ngradient, card-in-card layouts, magnetic cursor effects. They converge on the\nsame defaults because agents reach for training-data reflexes.\n\n**kiln** fixes this by separating permanent perceptual rules (human vision\ndoesn't change) from temporal taste (what's saturated right now). It ships six\nmaterial kits with distinct personalities, anti-pattern detection, and a\nquarterly-versioned zeitgeist file that tracks which aesthetics have been\nstrip-mined by AI generators.\n\n### The animation that doesn't feel right\n\nAgents add animation the way they add comments: everywhere, uniformly,\nwithout judgment. `transition: all`, scale(0) entrances, ease-in on\nexits, motion on keyboard actions that fire a hundred times a day.\n\n**animate-it** is the implementation protocol: gate (should this animate\nat all?), classify the motion type, specify easing/duration/tool, then\ncode against the rules that separate \"has animation\" from \"feels right.\"\nKiln decides what should move and why; animate-it makes it move\ncorrectly.\n\n### The wall of markdown\n\nAgents produce 300-line markdown plans nobody reads. Comparisons get\nstacked instead of side-by-side, timelines become bullet lists, and\nanything interactive gets described instead of built. The fix isn't\n\"always output HTML\" -- generated HTML has its own failure modes:\nslop styling, lost git diffs, dead-end editors.\n\n**htmlize** is the artifact protocol: gate (does this deliverable beat\nmarkdown?), shape (which artifact pattern), build (self-contained\nsingle file, calm typography, safe DOM), check (the review rubric).\nExports go through the clipboard; markdown stays the source of truth.\n\n### The first answer that sticks\n\nClaude picks the first reasonable approach and builds on it. For most\nproblems, that's fine. For decisions that are expensive to reverse --\narchitecture, naming, strategy -- it produces competent, forgettable\noutput. The real cost isn't a bad answer. It's never seeing the better\none because you stopped looking.\n\n**parallax** forces a structured pause: frame the decision, generate 3-5\napproaches that differ on fundamental assumptions (not implementation\ndetails), surface the traps, commit to one with explicit trade-offs.\nInspired by [ADHD](https://github.com/UditAkhourii/adhd)'s divergent\nideation work, but designed as a lightweight protocol -- not an\norchestration harness. One context, zero infrastructure.\n\n### The workflow you can't repeat\n\nYou spend 40 minutes getting an agent to do something perfectly -- tool\nsequences, corrections, domain knowledge discovered along the way. Next week\nyou need the same workflow and start from scratch.\n\n**skillize** crystallizes a session into a reusable skill. It surveys the\nconversation, classifies corrections by durability (permanent rule vs one-off\nfix), extracts the repeatable pattern, and drafts a SKILL.md. The workflow is\npreserved, not just the output.\n\n### Skills that can't travel alone\n\nYou build a skill that depends on another skill's scripts. You share the\n`.skill` file and the recipient's eval pipeline silently breaks because the\ndependency isn't there.\n\n**skillbun** resolves inter-skill dependencies, renames to avoid collisions\nwith independently installed copies, and packages everything into a single\n`.skill` archive. One file, one install.\n\n## Skills\n\n- **[kiln](skills/kiln/SKILL.md)** -- Precision UX design instrument.\n  Layered architecture: perceptual principles -\u003e composable patterns -\u003e\n  parameterized material palettes. Six curated kits. Anti-AI-slop detection.\n  Generates and audits interfaces.\n- **[animate-it](skills/animate-it/SKILL.md)** -- Animation implementation\n  protocol. Gate -\u003e classify -\u003e specify -\u003e code -\u003e check. Custom easing\n  curves, duration tables, tool selection, and a review rubric for\n  animation code that feels wrong.\n- **[htmlize](skills/htmlize/SKILL.md)** -- HTML artifact protocol.\n  Gate -\u003e shape -\u003e build -\u003e check. Turns agent deliverables into\n  self-contained HTML artifacts when they beat markdown -- and says no\n  when they don't. Starter templates, a build-time diagram pipeline\n  (graphviz / d2 / mermaid), clipboard exports, markdown as source of\n  truth.\n- **[parallax](skills/parallax/SKILL.md)** -- Multi-perspective decision\n  protocol. Frame -\u003e spread -\u003e commit. Forces genuinely different alternatives\n  before choosing. Trap detection as a first-class operation.\n- **[skillize](skills/skillize/SKILL.md)** -- Session-to-skill crystallizer.\n  Extracts repeatable workflows from agent sessions, classifies corrections,\n  drafts SKILL.md with references. Optional eval pipeline.\n- **[skillbun](skills/skillbun/SKILL.md)** -- Skill bundler. Resolves\n  dependencies, renames to avoid collisions, packages into `.skill` archives.\n\n## Repo structure\n\n```\n.\n├── .claude-plugin/           # plugin manifest for skills.sh\n│   └── plugin.json\n├── scripts/                  # repo utilities\n│   ├── link-skills.sh        #   symlink all skills to ~/.claude/skills/\n│   └── list-skills.sh        #   list all SKILL.md files\n├── skills/                   # flat -- each skill is self-contained\n│   ├── kiln/\n│   ├── animate-it/\n│   ├── htmlize/\n│   ├── parallax/\n│   ├── skillize/\n│   └── skillbun/\n└── template/                 # starter SKILL.md\n    └── SKILL.md\n```\n\n## Creating a new skill\n\n1. Copy `template/SKILL.md` into a new directory under `skills/`\n2. Name the directory to match the `name:` field in frontmatter\n3. Write the instructions (keep SKILL.md under 500 lines)\n4. Add the skill to `.claude-plugin/plugin.json`\n5. Test: `./scripts/link-skills.sh` then invoke in your agent\n\nOr: use `/skillize` after completing a workflow to auto-extract one.\n\n## Contributing\n\n1. Fork the repo\n2. Create your skill under `skills/`\n3. Validate: `name` matches directory, description under 1024 chars\n4. Add to `.claude-plugin/plugin.json`\n5. Open a PR\n\n## License\n\n[Apache 2.0](LICENSE)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flroolle%2Fskills","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flroolle%2Fskills","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flroolle%2Fskills/lists"}