A personal collection of agent skills. Each skill is a reusable, step-by-step playbook that an AI agent can follow — setting up servers, wiring deploys, git workflows — so the know-how is captured once and executed consistently every time.
Skills live under skills/, grouped by category. Each folder holds one skill as a SKILL.md file — see the file itself for the full details of what it does.
| Category | Skill | What it does |
|---|---|---|
development |
create-linear-task | Create a Linear task from text, screenshots or links: pick project, milestone, labels and estimate from the project's recent issues, write a short professional title and description, split cross-repo work into sub-issues, and create only after the draft is confirmed. |
development |
implement-linear-task | Implement a Linear task end to end: fetch the issue, plan, work on the task branch (never main), verify, then commit/PR/report back to Linear after confirmation. |
development |
review-linear-task | Review a Linear task's implementation against its requirements and end with a Complete/Incomplete/Needs-changes verdict; can post the review to Linear and merge after confirmation. |
devops |
setup-new-server | Harden a fresh Ubuntu/Debian VPS: sudo user, SSH keys and non-standard port, UFW, fail2ban, unattended upgrades, sysctl hardening. |
devops |
connect-repo-to-server | Give a server access to a private GitHub repo via a per-repo deploy key, then clone it. Only manual step is pasting the key into GitHub. |
frontend |
huzaifa | Huzaifa's personal UI taste for webapps: big heading + muted subtitle for section titles, and light/dark mode from day one via theme tokens. |
general |
unslop | Remove AI-writing tells from prose using a catalog of 36 patterns, with scope rules and per-format calibration. |
general |
todo | Manage a local SQLite task list by talking to Claude: add tasks to projects, see what is due today or this week (plus overdue), start/finish tasks by name or ID. All changes go through scripts/todo.py. |
git |
github-commit | Stage and commit changes with a message matching the repo's existing style. No co-author trailers, no pushes unless asked. |
security |
webapp-security-audit | Audit a webapp repo for vulnerable dependencies, leaked secrets (including git history), and hardening issues. Ends with a Critical/Medium/Small report. |
security |
mobileapp-security-audit | Same audit for mobile apps (Flutter-first), plus mobile hardening checks and a Dart source review. |
Point your agent at this repo, or copy a skill folder into wherever it looks for skills (e.g. .claude/skills/ or ~/.claude/skills/ for Claude Code). The agent picks a skill up automatically when a task matches its description — in Claude Code you can also invoke one directly with /<skill-name>.
See CONTRIBUTING.md for how skills are organized and what a good skill looks like.