skills.rest
Miasin-Labs
Miasin-LabsOFFICIAL
Explore all skills by this author →
GitHub

run-skill-generator

Generates per-project run skills with verified driver scripts for launching and driving applications.

9|2|Updated May 3, 2026
Search Tags:#agent workflow#skill authoring#build automation#e2e testing#headless#run instructions#driver script
One-click install
npx skills add https://github.com/Miasin-Labs/jfc --skill run-skill-generator-miasin-labs
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: run-skill-generator
Source: https://github.com/Miasin-Labs/jfc/tree/main/crates/jfc-agents/builtin-skills/claude-2.1.167/frontmatter-skills/run-skill-generator
Command: npx skills add https://github.com/Miasin-Labs/jfc --skill run-skill-generator-miasin-labs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents often lack a reliable, verified way to build, launch, and interact with a project's application from a clean environment, and README instructions alone are frequently inaccurate or untestable headlessly. ## Core Features & Use Cases - Verified Run Skill Authoring: Produces a SKILL.md plus a committed driver script (driver.mjs, smoke.sh, or chromium-cli heredoc) under .claude/skills/run-<unit>/ so future agents can programmatically drive the app. - Harness-First Workflow: Requires actually launching the app in the container, taking screenshots for GUIs, and documenting only commands that were executed and worked. - Project-Type Patterns: Provides starting templates for web servers, CLIs, TUIs, Electron desktop apps, browser-driven apps, and libraries. - Use Case: An agent asked to "set up and run the billing app" uses this skill to build a tmux-wrapped REPL driver, verify a real user flow end-to-end, and commit a run-billing skill that any future agent can invoke. ## Quick Start Ask the agent to create a run skill for this project so it can be built, launched, and driven from a clean environment.

Frequently Asked Questions about run-skill-generator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a run skill for my project?▼

Invoke the run-skill-generator and point it at one deployable unit. It discovers build steps, launches the app in the container, builds a driver harness, and writes a SKILL.md under .claude/skills/run-<unit>/ containing only verified commands.

What is a driver script in a run skill?▼

A driver is committed code that lets an agent programmatically interact with the running app, such as a Playwright REPL for Electron, a tmux wrapper for TUIs, curl for servers, or a chromium-cli heredoc for web apps.

Does the run skill work for headless Linux containers?▼

Yes. The workflow assumes a headless Linux container and uses xvfb-run for GUI apps, headless chromium-cli for web apps, and tmux for terminal apps, documenting the exact apt-get packages needed.

Where should the generated run skill be placed?▼

Place it at <unit>/.claude/skills/run-<unit-name>/ where <unit> is one deployable app or service. Claude Code natively discovers skills in nested .claude/skills directories and auto-loads them when descriptions match.

Why does the skill require actually running the app first?▼

Running the app proves the instructions work and surfaces real errors for the Troubleshooting section. Without execution, the skill merely paraphrases the README, which the workflow explicitly forbids.

Related Skills

Explore More Software Engineering Skills→
MinhOmegaMinhOmega

gen-claude-md

Generate CLAUDE.md at the repository root from analyzed structure and commands.

Community
Intermediate
mmnavarrmmnavarr

test-browser

Runs end-to-end browser tests on pages affected by a PR or branch using agent-browser.

Community
Advanced
onchainengineeronchainengineer

tests

Enforce unit and integration test structure with bun x jest.

Community
Intermediate
hospitaljobsinhospitaljobsin
7

e2e-testing-patterns

Provide end-to-end testing patterns for Playwright and Cypress.

Official
Advanced