Skip to content

Agent skills for real engineering

Not vibe coding. Small, composable, model-agnostic skills you can adapt to your repo instead of letting a process framework take over.

Built for Claude Code

Skills depend on Claude Code's slash commands, /plugin install path, and tool surface (Agent, Monitor, Bash, SendMessage, TaskStop). They aren't drop-in for Cursor, Codex, Cline, or general LLM chat. ("Model-agnostic" above means any Claude model — Opus, Sonnet, Haiku — not any harness.) See Compatibility for what porting would look like.

Get started in 60 seconds How the loop fits together

What changes when you install this

  • Without these skills

    "Build me a feature for X."

    Agent dives in, invents file paths, half-implements three layers, opens a PR that touches forty files and misses the constraint you only mentioned in passing. You spend an hour reverse-engineering what it built before you can review it.

  • With these skills

    /zsl:grill-with-docs interviews you against the project's CONTEXT.md and ADRs until the plan is concrete. /zsl:to-prd writes it up. /zsl:to-issues slices it into vertical-slice sub-issues with explicit Blocked by graphs. /zsl:tdd-parallel fans the unblocked AFK slices out into worktrees, merges in wave order, and opens one consolidated integration PR.

You stay in the loop where it matters (the plan, the slicing, the review) and delegate the parts that are mechanical (the test scaffolding, the merging, the busywork of opening N PRs).

Install

/plugin marketplace add ZunoSmartLabs/zsl-superpowers
/plugin install zsl@zsl-superpowers

Skills surface as /zsl:<skill-name> once installed (/zsl:tdd, /zsl:triage, …). Run /zsl:setup-zsl-superpowers once per repo to teach the engineering skills your issue tracker, label vocabulary, doc layout, and ship style.

Quickstart →

Why these skills exist

We built these to fix specific failure modes we kept hitting with Claude Code, Codex, and other coding agents:

You hit this Reach for
The agent didn't do what I want /zsl:grill-me or /zsl:grill-with-docs before you start
The agent is way too verbose /zsl:grill-with-docs — builds shared language inline in CONTEXT.md and ADRs
The code doesn't work /zsl:tdd for red-green-refactor; /zsl:diagnose when the bug is real
We built a ball of mud /zsl:improve-codebase-architecture every few days
I'm lost in this code /zsl:zoom-out
Need to break a PRD into work /zsl:to-prd → /zsl:to-issues → /zsl:triage
Multiple slices ready to ship at once /zsl:tdd-parallel — one integration PR, not N

Where to next

Quickstart : Install, run one skill, see what changed. Five minutes.

The workflow : How the skills compose into one engineering loop, end to end.

Parallel TDD deep-dive : Why we built it, how the wave model works, what an integration PR looks like.

Skills : Every skill, what it does, when it activates.

FAQ : Compatibility, telemetry, namespacing, opt-in subset.