omnistack-agent skill
omnistack-agent installs four skills for Claude Code, Copilot, Cursor and Codex: debug, review, security, engineering
- 68 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skills give the agent instructions for debugging, review and security checks and run nothing themselves
- The installer writes skill files into the chosen project
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/omnistack-agentDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills give the agent instructions for debugging, review and security checks and run nothing themselves
- The installer writes skill files into the chosen project
This is third-party code. Review the repository files before installing.
What it does
omnistack-agent builds short skills and reference material for engineering tasks from one source. Four ready skills: omnistack-agent for general work across twelve roles, omnistack-debug for reproducing a bug, isolating its cause and verifying a small fix, omnistack-code-review for findings with severity, file, line and impact, and omnistack-security-review for trust boundaries and realistic attack paths. The roles are architect, full stack, mobile, backend, frontend, database administrator, DevOps, QA, security, code reviewer and others. Tools, permissions, model selection and delegation come from the host platform. The scripts/install.mjs installer on Node 18 or newer takes a platform of claude, copilot, cursor or codex and a skill name. With --dry-run it only shows where it would write files, it refuses conflicts and it does not replace your project guidance. You need no build and no npm dependencies, because the generated files are already in the repository. npm run demo runs three demonstrations: a cart total bug, an access check bug and a task filter. The README stresses that the checks verify the demonstrations and do not measure model quality. For ChatGPT, Gemini Gem, Windsurf and others there are Markdown adapters you paste in by hand.
Who it is for. Developers who want to give an assistant ready working frames for debugging and review without assembling their own instruction set.
Good fit when
- You need to reproduce a bug, find its cause and verify a small fix step by step
- A review of changes should return findings with severity, file and line
- Before release you want trust boundaries and realistic attack paths examined
- You want to preview the install with --dry-run and be sure nothing is overwritten
- The assistant runs in ChatGPT or Gemini and needs instructions for a Custom GPT or a Gem
Not a fit when
- You want a skill for a narrow stack such as Kubernetes or one framework: the set is general
- You need proven model quality gains: the author checked only three demonstrations with no model responses evaluated
- The host platform cannot load skills and you will not paste Markdown adapters by hand
- You want autonomous agent work with its own rights: the platform sets tools and permissions
Example request
Install the omnistack-debug skill into this project for Claude Code, show --dry-run first, then debug the failing cart testLimitations
- npm run demo checks the author's three demonstrations and does not evaluate model responses.
- Two recorded Codex review trials show an observed response and a file-read policy limitation, and they are not a comparative test.
- Check that the host actually loaded the skill: /omnistack-debug in Claude Code, and other hosts have their own controls.
- The skills do not replace your project guidance, and manually added adapters are merged with existing rules.
- The installer and demonstrations need Node 18 or newer.
How to disable. Safe removal is described in docs/platforms.md in the repository. By hand, delete the installed skill's folder from the project's skills directory and the adapter block from CLAUDE.md or AGENTS.md if you added one.
FAQ
Which skills does omnistack-agent include?
Four: omnistack-agent for general engineering work, omnistack-debug for debugging, omnistack-code-review for review with severity, file and line and omnistack-security-review for trust boundaries and attack paths. Each skill carries its own reference files.
How do I install a skill in Claude Code?
Clone the repository and run node scripts/install.mjs install --platform claude --project /path/to/your-project --skill omnistack-debug. Run the same command with --dry-run first to see the paths. Invoke it in Claude Code with /omnistack-debug.
What do the demonstrations in npm run demo do?
Three tasks are checked in a disposable workspace: a cart total bug, an access check bug and a task filter feature. For each, the initial check fails and the solution checks pass. The script does not evaluate model responses.
Does it work in ChatGPT or Gemini?
Yes, through Markdown adapters. For ChatGPT paste custom-gpt-instructions.md into Instructions, and for a Gemini Gem use gem-instructions.md, optionally attaching the reference bundle.
Which roles does the general skill cover?
Twelve roles: architect, full stack, mobile and backend developer, frontend, database administrator, DevOps, QA, security, code reviewer and others. The host platform sets tools and permissions.
Discussion
No questions or reviews yet.
Related
SkillsA 13-skill delivery workflow for coding agents: from idea to plan, review, security audit, tests and a shipped pull request
Eleven skills for Claude Code and Codex: test audit, git worktree cleanup, plan pages, screen recording transcripts and 1Password sign-in
A set of 116 skills, 44 agents and safety hooks for Claude Code, Cursor, Codex, Gemini CLI and nine more editors
Catalog of 46 versioned skills for code review, releases, CI and landing pages, installed as the lstack plugin or through npx.
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A set of skills from real work: npm supply-chain audit, CLI building, Obsidian plugin release and an AGENTS.md hierarchy
README in short
The README walks through one scenario: clone the repository, choose a platform, preview the install with --dry-run, install a skill, try a copyable prompt and run the checks. Next come three reproducible demonstrations, a table of adapters for other platforms and the twelve roles. The single-source build needs no npm dependencies.
README badge
Original SKILL.md
--- name: omnistack-debug description: "Diagnose a reported software failure, reproduce its cause, and verify a focused regression fix." --- <!-- GENERATED from core/ + workflows/ + knowledge/ — DO NOT EDIT — run: npm run build --> Read [shared engineering guidance](references/core.md) before applying the focused workflow below. # Diagnose and fix a reported failure Work from the user's reproduction, logs, relevant code and installed versions. Keep observations separate from competing explanations. Reproduce the failure with the smallest useful input when tools allow it; otherwise state what the supplied evidence establishes and what remains uncertain. Trace the input and state through the failing path before editing. Choose the next probe that distinguishes plausible causes; stop repeating a failed approach without new evidence. Fix the demonstrated cause with a focused change and verify the regression plus affected boundaries. Preserve unrelated code and user edits. Report the cause, affected path, correction and actual check results. Missing access or an unrun reproduction is a limitation, not a passing result. Do not invent logs or attribute a failure to an unobserved dependency. ## Reference modules Read only the reference relevant to this task; these files are packaged beside this skill. - [JavaScript Essentials](references/knowledge/languages/javascript.md) - [TypeScript Essentials](references/knowledge/languages/typescript.md) - [API Design](references/knowledge/backend/apis.md) - [Relational Databases](references/knowledge/databases/relational.md) - [CI/CD](references/knowledge/devops/ci-cd.md) - [Automated Testing](references/knowledge/testing/automated.md)