Skip to content
Agentmart
Submit a tool

steroid-engine

The autonomous execution orchestrator for Steroid-Workflow.

Skill by nzkbuild

IndexedMIT★ 3
Write a review

Chalk Labs

Need help deploying steroid-engine?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/nzkbuild-steroid-engine

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

The autonomous execution orchestrator for Steroid-Workflow. This skill reads .memory/changes/<feature>/plan.md and loops through the execution checklist using TDD, committing atomically per task, capturing learnings, and stopping only on completion or circuit breaker trip.

Use cases: testing, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/steroid-engine

What's inside

Skill name
steroid-engine
Description
The autonomous execution orchestrator for Steroid-Workflow. This skill reads .memory/changes/<feature>/plan.md and loops through the execution checklist using TDD, committing atomically per task, capturing learnings, and stopping only on completion or circuit breaker trip.
License
MIT
Instructions
508 lines in SKILL.md · view on GitHub

Agent Skills spec issues

  • body: 508 lines (spec recommends under 500)

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Steroid Engine (Autonomous Execution Orchestrator)

This skill autonomously executes the checklist in `.memory/changes/<feature>/plan.md`. It uses the Subagent-Driven Development model forked from `Steroid` (see `Steroid internal reference` and `Steroid internal reference`) combined with the autonomous loop pattern from Steroid (see `Steroid internal reference`).

## Governed Baseline

The live governed authority for this phase is:

- `Steroid runtime contract`
- `Steroid runtime contract`
- `Steroid runtime contract`
- `Steroid runtime contract`

In the live repo, this skill is the execution entry point for the governed `steroid-execution-engine`. The current runtime emits standalone live equivalents for `tasks_md` and `execution_receipts` at `.memory/changes/<feature>/tasks.md` and `.memory/changes/<feature>/execution.json`.
In the live repo, this skill is also one of the runtime surfaces for the governed `steroid-progress-memory`. `.memory/progress.md` is the durable append-only progress artifact.

If `.memory/changes/<feature>/prompt.json` exists, read it before the first task. Treat its assumptions, non-goals, continuation state, and recommended route as execution guardrails. They do not replace `plan.md` or `diagnosis.md`, but they explain why the current plan is shaped the way it is.

## The Circuit Breaker Mandate

All terminal commands executed by this skill or any sub-agent it dispatches MUST be wrapped in the physical Node.js circuit breaker:

```
node steroid-run.cjs '<command>'
```

Direct terminal commands (`npm install`, `npx jest`, `node script.js`, etc.) are strictly forbidden. The `steroid-run` wrapper physically tracks errors in `.memory/execution_state.json`. If the error count reaches 5, the wrapper hard-faults and prevents further execution. The AI has no ability to override this.

## Anti-Summarization Directive

NEVER summarize code. NEVER write comments like `// rest of the code here` or `// ...existing code...`. NEVER truncate file contents. If modifying a file, write the complete replacement content or use precise line-targeted edits.

Before marking any task as `[x]`, run the verification guard:

## .gitignore Protection (v5.0.1)

NEVER overwrite the project's `.gitignore`. The installer appends steroid entries (`.memory/`, `.steroid/`, `steroid-run.cjs`, `.agents/`). If you need to create or modify `.gitignore`, ALWAYS APPEND to the existing file — never replace it. The commit command will auto-restore missing entries, but prevention is better than cure.

## Protected Files (v5.0.2)

These files MUST NOT be replaced entirely. Always read first, then modify:

| File                              | Rule                                           |
| --------------------------------- | ---------------------------------------------- |
| `.gitignore`                      | APPEND only. Never replace.                    |
| `package.json`                    | Add/update fields. Never rewrite from scratch. |
| `tsconfig.json`                   | Modify options. Never replace.                 |
| `.env` / `.env.local`             | NEVER touch. Contains secrets.                 |
| `next.config.*` / `vite.config.*` | Read before modifying.                         |
| Any file in `.memory/`            | Read-only during engine phase.                 |

If you must create one of these files for a fresh project, ensure existing content (if any) is preserved.

### Anti-Deletion Guard (v5.5.0)

When modifying an existing file, you MUST preserve all existing:

- Exported functions and components
- UI elements and layouts
- Route definitions
- Configuration values

You may only delete existing code if the spec or plan **explicitly** says to remove it. "Refactoring" is not a valid reason to delete working features. If in doubt, ADD alongside existing code rather than REPLACING it.

```
node steroid-run.cjs verify <path/to/file> --min-lines=<expected>
```

If the file is shorter than expected, the verify command will exit with an error, blocking the task from completion.

## Scaffold Safety (v5.6.0)

NEVER run scaffold commands (`npm create`, `npx create-*`, `npm init`, `yarn create`, `pnpm create`) targeting the project ROOT directory (`.`).

The circuit breaker will physically block these commands. If you attempt `node steroid-run.cjs 'npm create vite . ...'`, the circuit breaker will refuse execution and suggest the safe alternative.

**Safe pattern (v6.0.0 — uses shell-free fs-\* subcommands):**

1. Scaffold into a temp subdirectory:
   `node steroid-run.cjs 'npx create-next-app@latest .steroid-scaffold-tmp --typescript ...'`
2. Copy scaffolded files into root:
   `node steroid-run.cjs fs-cp .steroid-scaffold-tmp .`
3. Remove temp directory:
   `node steroid-run.cjs fs-rm .steroid-scaffold-tmp`
4. Install dependencies:
   `node steroid-run.cjs 'npm install'`
5. Rescan to update memory:
   `node steroid-run.cjs scan <feature> --force`

**Why:** Scaffold tools like `create-vite` prompt "Remove existing files and continue" when the target directory is non-empty. If the AI selects this option (intentionally or accidentally), it deletes `.git/`, `.memory/`, `steroid-run.cjs`, and all infrastructure — destroying the circuit breaker itself and making recovery impossible.

## Phase Gate (Physical Enforcement)

Before doing anything, run the gate check:

```
node steroid-run.cjs gate engine <feature>
```

If this command fails, STOP. The architecture phase is not complete.

## Pre-Execution Setup

Before starting the checklist, create a rollback safety point:

```
node steroid-run.cjs git-init
```

If the build goes wrong, the user can recover with `git reset --hard`.

Also inspect whether this feature is coming from:

- `plan.md` for the normal build/refactor/migrate/document path
- `diagnosis.md` for the targeted fix path

If the feature came from a `split-work` route, complete one clearly scoped sub-problem or story at a time instead of mixing separate intents in the same execution burst.

## Live Execution Artifacts

From nzkbuild/steroid-workflow, MIT.

More from nzkbuild/steroid-workflow

Signals

Stars
3
Forks
0
Open issues
0
Last commit (this folder)
Mar 26, 2026 (7 months ago)
Last release
Mar 16, 2026 (7 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from nzkbuild

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get steroid-engine running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying steroid-engine?

CallGet help
Cite this page

steroid-engine skill. Agent Mart. https://agentmart.dev/skills/nzkbuild-steroid-engine (last updated 2026-10-24)