Minimal reproduction

Skill yeaight7/agent-powerups/skills/minimal-reproduction

Use when a bug must be isolated from a large application into a standalone, runnable script or single test case.From its SKILL.md

Install

npx -y skills add yeaight7/agent-powerups --skill minimal-reproduction

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, 611 tokens by cl100k_base, as published. Nobody here has run it

Purpose

You cannot reliably fix what you cannot reliably reproduce in isolation. Apply the subtraction method until a single file deterministically triggers the exact reported error.

When to Use

A bug is tangled in app/UI/network layers and hard to debug in place
A failure must be handed to another engineer, team, or upstream project
Before bisecting or stress-testing, to get a fast deterministic signal

Inputs

The failing code path and the error it produces
Sample data that triggers the failure, if any

Workflow — the Subtraction Method

Start with the failure. Take the code path that fails.
Remove the UI/Network. If the bug is reported via a web request, write a script that calls the internal controller directly.

Mock dependencies. If the bug doesn't require the database, mock it. If it doesn't require the third-party API, mock it. Template:

// repro.js — minimal harness, no framework
const { failingFunction } = require("./src/module");

const fakeDb = { query: async () => [{ id: 1, value: null }] }; // smallest stub that triggers it

failingFunction(fakeDb, { payloadKey: "trigger-value" })
  .then(() => { console.log("NO REPRO"); process.exit(0); })
  .catch((err) => { console.error("REPRO:", err.message); process.exit(1); });

(Exit 0 on no-repro / non-zero on repro also makes the script directly usable by git bisect run.)

Prune data. If the bug fails on a 10MB JSON payload, binary search the payload down to the exact 2 keys that trigger the failure.
Final output. The result must be a single file — a standalone script or one test case — that relies on ZERO external state, can be run with a single command, and deterministically outputs the exact error reported.

Output

One standalone file plus the single command that runs it (e.g., node repro.js)
The exact error output, proving the reproduction matches the report

Verification

Repro runs with one command and no external state (no DB, network, or special env)
Error produced matches the reported error exactly
Repro is deterministic across at least 3 consecutive runs
Every remaining line is necessary — removing any piece makes the bug disappear

Failure Modes

Partial isolation — the "repro" still needs a live database or API; it will rot immediately.
Wrong error — the script fails, but with a different error than reported; that is a different bug.
Over-pruning — cutting until the bug vanishes and shipping the last version untested; re-run after every cut.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most root cause skills give in 611 tokens

Counted across 120 of the 151 authors here whose files we hold, read 2026-09-06

Add stack trace instrumentation when manual tracing failsin 8 of 120, across 5 files
Trace backward through the call chainin 8 of 120, across 5 files
Use console.error in tests instead of loggerin 8 of 120, across 5 files
Capture the full call chain with new Error().stackin 8 of 120, across 5 files
Fix the root cause, not the symptomin 7 of 120
Log before the dangerous operation, not after it failsin 7 of 120, across 4 files
Log directory, cwd, environment variables, and stackin 7 of 120, across 4 files
Add validation at each layer as defense-in-depthin 6 of 120, across 3 files
Run tests and grep for the debug outputin 6 of 120, across 3 files
Analyze stack traces for test file names and line numbersin 6 of 120, across 3 files
Fix at the source, not the symptomin 5 of 120, across 2 files
Assign every action item an ownerin 5 of 120

Said here and by no other author read

Apply the subtraction method until one file triggers the error
Remove UI and network layers by calling internal code directly
Mock dependencies the bug does not require
Binary search the data down to the triggering keys
Exit non-zero on repro and zero on no-repro
Make the repro runnable with one command

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.