Skip to content
Back to skills

Qstack Build The Lever

ASecurity

Build the smallest rerunnable tool that makes non-trivial work repeatable, reviewable, and safe to verify.

  • 7 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 5, 2026
ai-agentsrustgit

Works with

  • cursor

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 5, 2026

npx -y skills add hani-q/qstack --skill qstack-build-the-lever --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Qstack Build The Lever?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Qstack Build The Lever
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/hani-q-qstack-build-the-lever/badge)](https://www.skillsdirectory.com/skills/hani-q-qstack-build-the-lever)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: qstack-build-the-lever
description: Build the smallest rerunnable tool that makes non-trivial work repeatable, reviewable, and safe to verify.
disable-model-invocation: true
license: MIT
---

# Build the lever

When authorized work is not a couple of obvious edits, build the smallest tool
that performs or proves it instead of repeating the work by hand.

## Build one rerunnable artifact

1. Do one unit manually to learn the real recipe.
2. Choose the smallest fitting artifact: a script, codemod, generator, query,
   check, or a single immutable recipe for delegated work. Build a tool, not a
   framework.
3. Make reruns safe. Detect current state, avoid duplicate effects, expose
   failures, and support a dry run when mutations carry meaningful risk.
4. Run the tool against the first unit and diff its result against the manual
   version. Resolve any mismatch before applying it broadly.
5. Use the tool for the remaining scope and keep its output reviewable.
6. Prove the produced artifacts directly. A deterministic check is part of the
   lever when it is what makes the work trustworthy.

A deterministic lever beats several people or agents applying the same recipe
by hand. If delegation remains useful, give every delegate the same recipe,
verification contract, and write fence. Keep that contract outside their write
scope.

Applying this skill to authorized non-trivial work produces a rerunnable file.
For a read-only request, propose the file and do not create it. Keep every write
inside the user's scope. Invocation alone never authorizes a commit, push,
publication, deployment, or external message.

Adapted from Lauren Tan's PStack
[`principle-build-the-lever`](https://github.com/cursor/plugins/blob/60c641e4fad674784b30abcf9f8915dea39df38d/pstack/skills/principle-build-the-lever/SKILL.md)
at commit `60c641e4fad674784b30abcf9f8915dea39df38d` under the MIT License.
See [third-party notices](../../THIRD_PARTY_NOTICES.md).

Files in this skill

  • SKILL.md1.9 KB
  • agents/openai.yaml264 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.