- Home
- Skills
- Agents & Orchestration
- agent-setup
Agent Setup
Scaffold a project-local team of specialized subagents grounded in your repository's stack and conventions.
$2
Works with the AI Agents you already use
agent-setup
Example session with this skill installed
/agent-setup in this repo. It's a TypeScript Next.js app with Prisma and Vitest.
- Read your context and instructions
- Compiled the agent-setup
I have analyzed the repository structure. Based on the presence of package.json, prisma/schema.prisma, and vitest.config.ts, I am scaffolding a specialized agent team for the web-portal project.
Created the following agents in ./.claude/agents/:
web-portal-dev.md: Implements features using Next.js conventions and Vitest.web-portal-reviewer.md: Read-only agent for diff analysis and PR readiness.web-portal-testing.md: Specialized for writing and maintaining Vitest suites.web-portal-data-expert.md: Grounded in Prisma schema for migrations and DB logic.
Project Agent Team
| Role | File | Capabilities |
|---|---|---|
| Developer | web-portal-dev.md | Feature implementation, Next.js API routes, UI components. |
| Reviewer | web-portal-reviewer.md | Read-only code review, linting check, logic validation. |
| Testing | web-portal-testing.md | Unit and integration testing via npm run test. |
| Data Expert | web-portal-data-expert.md | Prisma migrations, schema optimization, seed scripts. |
Next steps
- Review the generated
.mdfiles in.claude/agents/to ensure the specific commands match your local aliases. - Invoke a specific agent by name, e.g.,
@web-portal-reviewer, when you are ready to check your staged changes. - Add these files to your
.gitignoreif you prefer them to stay local, or commit them to share the agent team with your collaborators.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop manually configuring project-specific agent instructions for every new repository. This skill automates the creation of a localized agent team tailored to your specific codebase, stack, and architectural patterns. It prevents configuration drift by grounding every agent in your project's actual source code, directory structures, and CI conventions.
The skill avoids the common mistake of polluting global agent configurations with project-specific paths or internal service names. By writing directly to your repository's local configuration directory, it ensures that your team's specialized knowledge—like specific migration workflows or test suites—stays within the project context where it belongs.
What it does
- Repo grounding resolves the absolute project root to ensure all generated agents operate within the correct local scope.
- Stack investigation analyzes
package.json,go.mod,Cargo.toml, and other manifest files to detect languages and build tools. - Role derivation determines the optimal set of subagents, such as developers, reviewers, or architects, based on project complexity.
- Local scaffolding writes specialized Markdown agent definitions directly to the project's
.claude/agents/directory. - Conflict prevention checks for existing agent slugs and prompts for refresh or skip to protect manual edits.
How it works
- Ground the target by resolving the git root or current working directory to scope all subsequent file operations.
- Investigate the repo via targeted greps and file reads to identify testing frameworks, database migrations, and monorepo structures.
- Decide the role set by selecting roles like
dev,reviewer, ordata-expertbased on the signals found during investigation. - Write the agents as individual Markdown files with specific tool access and instructions grounded in the repo's real commands.
Frameworks & tools
This skill works with any language detected in the repo, including JavaScript/TypeScript, Go, Rust, Python, and Java. It integrates with standard build tools and version control systems using git and bash.
Why this beats prompting it yourself
Writing project-specific instructions manually is repetitive and prone to path errors. This skill ensures your agents know the exact test commands and module boundaries of your current repo without you having to explain them every time you start a new session.
Use cases
- Scaffold a pair of dev and reviewer agents for a new microservice.
- Generate a specialized data-expert agent for a repo containing complex SQL migrations.
- Create an architect agent for large monorepos to handle design planning.
- Refresh existing agent instructions after a major refactor or stack change.
Known limitations
The skill requires a local .claude/agents/ directory structure to exist or be created within the project root. It does not automatically commit generated files to version control.
Requires git or a standard project structure to resolve root. Does not stage or commit files. Limited to local project scope.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 18 days ago
Requires