Skip to content
RL22Public

About

Self-authored Agent Skills for Claude Code, Codex, Cursor and OpenCode: design, content, agent ops and automation.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Skills

23 self-authored skills that teach your coding agent to design, write, ship and run its own ops.

Plain Agent Skills folders: one SKILL.md each, no plugin runtime. They work in Claude Code, Codex, Cursor, OpenCode and any agent that reads the standard.

License: MIT Skills: 23 Agent Skills standard

Skills • Quickstart • Configuration • License

Hand-drawn map of the repo: your agent branches into Design, Content, Agent ops and Automation skills

Drawn by sketch, one of the skills below, from this short JSON spec.

Quickstart

Clone once, then link the skills you want into your agent's skills folder:

git clone https://github.com/RL22/skills.git
ln -s "$PWD/skills/sketch" ~/.claude/skills/sketch

Use ~/.agents/skills/ or your agent's own skills directory in place of ~/.claude/skills/ as needed. Invoke a skill by name (/sketch flow …) or just describe the task; the agent picks it from its description.

Tip

Link several at once: for s in sketch og-image readme; do ln -s "$PWD/skills/$s" ~/.claude/skills/$s; done

Skills

1 · Design and visuals

Skill What it does
sketch Hand-drawn, lo-fi UX deliverables from a JSON spec: flows, screens, component sheets, sitemaps, journey maps, traces of live pages, responsive sets.
og-image Open Graph / social preview cards (1200×630) from an atomic design system with @vercel/og.
programmatic-brand-assets Logos, favicons, PWA and app icons from SVG, Canvas, p5.js or deterministic geometry.
img-gen Create and edit images: photos, logos, banners.
img-brandkit Brand-kit boards (3×3 and 2×3 identity grids, logo systems, visual decks), prompted and generated.
create-visuals Design rules and motion patterns for polished Remotion videos and animated graphics.
create-thumbnail YouTube thumbnail candidates: cut the subject out with rembg, then composite it.
virtual-staging Real-estate virtual staging that keeps the original architecture intact across angles.

2 · Content and writing

Skill What it does
readme READMEs written like landing pages in GitHub Flavored Markdown, plus an audit script that scores one.
writing-for-humans Draft, revise or review clear, specific, human-sounding writing.
content-ideas Scrape the creators you track, score what's performing, and turn it into ideas.
content-outline A structured script outline from a YouTube URL, article, markdown file or raw idea.
video-edit Raw footage to a published video: transcription, FFmpeg cuts, Remotion graphics.
elg-engine Social posts from git diffs, PRDs and changelogs, written from five perspectives.

3 · Agent ops

Skill What it does
model-delegation Route work across the Gemini, GPT and Claude CLIs: bulk reads, live search, media, second-opinion reviews.
map-graph Plan complex tasks as agent graphs (DAGs) with model tiers, state schemas and guardrails.
eval-engine Design and run agent evaluations, from binary verifiers to trace-mining improvement loops.
failure-ledger Turn failed runs into root causes and standing rules in AGENTS.md.
absorb Pull patterns out of a video, article or thread and apply them to your own project.
skillvault Keep one portable skills library as the source of truth, and vet new skills before adding them.

4 · Automation

Skill What it does
gws-gmail-cleanup Audit Gmail storage with the Google Workspace CLI and clean it with rules you approve.
job-search-operator Score job descriptions, tailor applications, and prepare outreach and interviews.
linkedin-scout Find people, jobs, companies and posts from your own logged-in LinkedIn session.

Dependencies between skills

Some skills call scripts in a sibling skill by relative path, so install them together:

  • img-brandkit uses model-delegation.
  • job-search-operator uses linkedin-scout (override with JSB_LINKEDIN_SCOUT_CLI).

linkedin-scout also needs capt-chrome-agent, which is not in this repo. Put it next to linkedin-scout in the same skills folder, or set CAPT_CHROME_AGENT_DIR. absorb uses it too, but only for dynamic or gated pages.

Configuration

Most skills need nothing beyond the agent. These need a one-time setup:

sketch: Node 20+ and a headless Chromium
cd skills/sketch
npm install && npx playwright-core install chromium
content-outline, video-edit, create-thumbnail, linkedin-scout: working directories

content-outline, video-edit and create-thumbnail read SPRINTZ_CONTENT_DIR; linkedin-scout reads SPRINTZ_JOB_SEARCH_DIR. Both default to folders under ~/Sprintz/. Point them at your own directories:

export SPRINTZ_CONTENT_DIR="$HOME/content"
export SPRINTZ_JOB_SEARCH_DIR="$HOME/job-search"

Important

linkedin-scout automates a logged-in LinkedIn session. Review LinkedIn's terms and use it only on your own account and data. All fixtures in this repo are fictional.

License

MIT; see LICENSE. A skill's own license: frontmatter takes precedence: elg-engine is Apache-2.0. sketch bundles rough.js (MIT) and three SIL OFL fonts, with their licenses in sketch/scripts/vendor/.

About

Self-authored Agent Skills for Claude Code, Codex, Cursor and OpenCode: design, content, agent ops and automation.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages