9 skills to automate dev routine

Reels hype "9 skills that automate your dev routine" — except none of those names turn up in a search. Here are 9 real tools for the same jobs: with repos, stars and nothing made up 🔧

What it is and who it's for

A set of skills, agents and open-source tools that cover dev routine: spec and plan, a process with TDD, test generation, refactoring, dependencies, API contracts. All of it checked by hand: official repos, live releases, real star counts as of July 2026.

Who it's for: the vibe coder on Claude Code (or a compatible agent) who is tired of repeating the same things by hand and wants the agent to run the process itself — from requirements to tests.

Key capabilities

1. Spec Kit (121k★, MIT) — spec → plan → tasks → code via the /speckit.specify, /speckit.plan, /speckit.tasks, /speckit.implement commands. An official GitHub project, works with Claude Code, Copilot, Cursor and dozens of other agents.
2. superpowers (254k★, MIT) — a whole methodology: brainstorming → plan → TDD implementation → systematic debugging → code review. Release v6.1.1 landed on 2 July 2026 — the project is alive.
3. anthropics/skills (161k★) — Anthropic's official Agent Skills repo: the reference for how skills are built, plus ready-made skills from documents to web-app testing.
4. wshobson/agents (37.9k★, MIT) — a marketplace of 92 plugins: 199 agents and 162 skills for architecture, languages, infrastructure, security and review.
5. awesome-agent-skills (28k★, MIT) — a catalog of 1000+ skills from official teams and the community: look for a niche skill here, not in reels.
6. Renovate (22k★, AGPL) — automatic dependency updates as scheduled PRs; developed by Mend.io, the de facto standard for this job.
7. Qodo Cover (5.5k★, AGPL) — AI test generation that iterates on code coverage: it writes them, runs them and keeps only the passing ones.
8. OpenRewrite (3.6k★, Apache-2.0) — bulk refactoring via ready-made "recipes": framework migrations, security fixes, consistent style.
9. Spectral (3.2k★, Apache-2.0) — an API contract linter: OpenAPI/JSON/YAML against your own style guides, so the contract doesn't drift from reality.

As for the reel's "performance scanner" and "error handling standardizer" — there's no canonical ready tool for those; for your own conventions it's safer to build your own skill (see the skill-creator card).

Get started in 5 minutes

1. Install superpowers in Claude Code: /plugin install superpowers@claude-plugins-official — then ask: "take feature X using the superpowers methodology".
2. Want the spec-driven approach — install Spec Kit: uv tool install specify-cli --from git+https://github.com/github/spec-kit.git, then specify init my-project and the /speckit.* commands in your agent (needs uv and Python 3.11+).
3. Hook up the agent marketplace: /plugin marketplace add wshobson/agents — and install only the plugins you need.
4. Turn on Renovate for the repo (GitHub app or self-hosted) — dependency update PRs will start arriving on their own.

When to use it, when not to

✅ Take superpowers if the agent starts slinging code with no plan and no tests — the methodology forces it to think before coding. ✅ Take Spec Kit if features are large and context gets lost: the spec becomes an executable source of truth. ✅ Take Renovate + Qodo Cover if your routine is "update dependencies and don't break anything": an update PR plus tests as insurance. ❌ Don't take the whole set at once — start with one: on small edits the spec-and-TDD ritual takes longer than the task itself. ❌ Don't take OpenRewrite outside the JVM world — its recipes are strongest in Java/Kotlin/Spring; for JS/TS it's simpler to ask the agent directly with tests as a safety net. ❌ Don't hunt for "the skills from the reel" in random marketplaces — instead of the non-existent "Auto Refactor Pro" browse the awesome-agent-skills catalog.

The honest price

Everything here is free open source: MIT (superpowers, Spec Kit, wshobson/agents, the awesome catalog), Apache-2.0 (OpenRewrite, Spectral), AGPL-3.0 (Renovate, Qodo Cover — check the license if you're embedding their code in your own product). The hidden price is tokens: superpowers, Spec Kit and Qodo Cover all drive an LLM, so your API or agent-subscription bill grows in proportion to your ambitions. Tag: free / oss.

Gotchas

• The names from reels ("Rapid Spec Builder", "Auto Refactor Pro", "API Contract Guardian" and friends) aren't products, they're pretty names for someone's prompts. Don't install same-named knockoffs: a third-party skill is someone else's instructions inside your agent (hello, prompt injection).
• superpowers imposes a process: brainstorming and TDD for every little thing — overkill for trivia.
• Spec Kit needs uv and Python 3.11+, and the spec only pays off if you keep it honestly rather than as a formality.
• Qodo Cover updates rarely (last push — April 2026) and can generate box-ticking tests — review what they actually verify.
• Read any skill from the catalog before installing it (the SKILL.md file) — like a README on an npm package.
• Stars ≠ quality for your stack: half the value is in tuning it to your own conventions.

🤖 Prompt accelerator

"I have superpowers and Spec Kit installed. Take feature : 1) run brainstorming and lock the requirements into a spec via /speckit.specify; 2) draft a plan and break it into tasks (/speckit.plan, /speckit.tasks); 3) implement it TDD-style: test first, then code, run the tests after every step; 4) at the end review the changed files and show what from the spec is covered and what isn't. Don't write code before the plan is approved."