advanced⏱70 minLección 10 de 10

Real-World Agent Projects

Apply agentic AI techniques to code generation, code review, refactoring, testing, documentation, and DevOps automation.

Real-World Agent Projects

Overview

This capstone lesson synthesizes everything you have learned into practical, real-world projects. Each project demonstrates how agentic AI techniques apply to common software engineering tasks.

💡 Tip: Use Ctrl + Mouse Wheel to zoom, or click the fullscreen icon

ℹ️Note

These projects are designed to be run with OpenCode or any agent framework. Each project includes the configuration, skill definitions, and expected outputs needed for successful execution.

Project 1: Automated Code Generation

Generate a complete REST API module from a specification using an agent.

💡 Tip: Use Ctrl + Mouse Wheel to zoom, or click the fullscreen icon

yaml

# project-configs/blog-api/skill.yaml
name: blog-api-generator
description: Generate a complete blog REST API from specifications
version: "1.0.0"
instructions:
  - path: instructions/generate-models.md
  - path: instructions/generate-routes.md
  - path: instructions/generate-tests.md
tools:
  required:
    - read
    - write
    - glob
    - bash
resources:
  - path: templates/fastapi-base.py.hbs
  - path: templates/sqlalchemy-model.py.hbs
  - path: templates/pytest-fixtures.py.hbs
autoload:
  enabled: false

python

# Example: Agent generating a FastAPI blog module
GENERATED_MODULE = """
# models.py
from sqlalchemy import Column, Integer, String, Text, DateTime, ForeignKey
from sqlalchemy.sql import func
from database import Base

class User(Base):
    __tablename__ = "users"

    id = Column(Integer, primary_key=True, index=True)
    username = Column(String(50), unique=True, nullable=False)
    email = Column(String(120), unique=True, nullable=False)
    created_at = Column(DateTime(timezone=True), server_default=func.now())

class Post(Base):
    __tablename__ = "posts"

    id = Column(Integer, primary_key=True, index=True)
    title = Column(String(200), nullable=False)
    content = Column(Text, nullable=False)
    author_id = Column(Integer, ForeignKey("users.id"), nullable=False)
    created_at = Column(DateTime(timezone=True), server_default=func.now())
    updated_at = Column(DateTime(timezone=True), onupdate=func.now())

# schemas.py
from pydantic import BaseModel, EmailStr
from datetime import datetime
from typing import Optional

class UserCreate(BaseModel):
    username: str
    email: EmailStr

class UserResponse(BaseModel):
    id: int
    username: str
    email: str
    created_at: datetime

class PostCreate(BaseModel):
    title: str
    content: str

class PostResponse(BaseModel):
    id: int
    title: str
    content: str
    author_id: int
    created_at: datetime
    updated_at: Optional[datetime] = None

# routes.py
from fastapi import APIRouter, Depends, HTTPException, status
from sqlalchemy.orm import Session
from . import models, schemas
from .database import get_db

router = APIRouter(prefix="/api/v1", tags=["blog"])

@router.post("/users/", response_model=schemas.UserResponse, status_code=201)
def create_user(user: schemas.UserCreate, db: Session = Depends(get_db)):
    db_user = models.User(**user.model_dump())
    db.add(db_user)
    db.commit()
    db.refresh(db_user)
    return db_user

@router.get("/posts/", response_model=list[schemas.PostResponse])
def list_posts(skip: int = 0, limit: int = 10, db: Session = Depends(get_db)):
    return db.query(models.Post).offset(skip).limit(limit).all()

@router.post("/posts/", response_model=schemas.PostResponse, status_code=201)
def create_post(post: schemas.PostCreate, db: Session = Depends(get_db)):
    db_post = models.Post(**post.model_dump())
    db.add(db_post)
    db.commit()
    db.refresh(db_post)
    return db_post

@router.get("/posts/{post_id}", response_model=schemas.PostResponse)
def get_post(post_id: int, db: Session = Depends(get_db)):
    post = db.query(models.Post).filter(models.Post.id == post_id).first()
    if not post:
        raise HTTPException(status_code=404, detail="Post not found")
    return post
"""

Project 2: Intelligent Code Review

An agent that reviews pull requests for bugs, security issues, and style violations.

json

{
  "skills": {
    "pr-reviewer": {
      "manifest": "skills/pr-reviewer/skill.yaml",
      "autoLoad": true,
      "matchPattern": "review pull request|review PR|#\\d+"
    }
  },
  "agents": {
    "reviewer": {
      "model": "claude-sonnet-4-20250514",
      "description": "Code review specialist",
      "prompt": "You are a senior code reviewer. Review code for: security vulnerabilities, performance issues, maintainability problems, and style violations.",
      "constraints": {
        "allowedTools": ["read", "grep", "glob", "bash"],
        "deniedTools": ["write", "edit"]
      }
    }
  }
}

python

class CodeReviewAgent:
    def __init__(self, agent):
        self.agent = agent
        self.findings = []

    async def review_pr(self, pr_number, repo_path):
        # Get the diff
        diff = await self.agent.bash(f"cd {repo_path} && git diff main...HEAD")
        changed_files = await self.agent.bash(
            f"cd {repo_path} && git diff --name-only main...HEAD"
        )

        files = changed_files.strip().split("\n")
        for file in files[:10]:
            if file.endswith((".py", ".ts", ".js", ".tsx", ".jsx")):
                content = await self.agent.read(f"{repo_path}/{file}")
                file_review = await self._review_file(file, content)
                self.findings.extend(file_review)

        return self._generate_report()

    async def _review_file(self, file_path, content):
        findings = []
        lines = content.split("\n")

        for i, line in enumerate(lines, 1):
            # Security checks
            if "eval(" in line or "exec(" in line:
                findings.append({
                    "severity": "critical",
                    "file": file_path,
                    "line": i,
                    "type": "code_injection",
                    "description": f"Dangerous function usage at line {i}"
                })

            # Performance checks
            if "for " in line and any(
                nested_line.strip().startswith("for ")
                for nested_line in lines[i:i+5]
            ):
                findings.append({
                    "severity": "major",
                    "file": file_path,
                    "line": i,
                    "type": "nested_loop",
                    "description": "Potential O(n^2) nested loop detected"
                })

            # Hardcoded secrets
            if "password" in line.lower() and "=" in line:
                findings.append({
                    "severity": "critical",
                    "file": file_path,
                    "line": i,
                    "type": "hardcoded_secret",
                    "description": "Possible hardcoded credential"
                })

            # TODO/FIXME tracking
            if "TODO" in line or "FIXME" in line:
                findings.append({
                    "severity": "info",
                    "file": file_path,
                    "line": i,
                    "type": "todo",
                    "description": line.strip()
                })

        return findings

    def _generate_report(self):
        by_severity = {}
        for f in self.findings:
            sev = f["severity"]
            if sev not in by_severity:
                by_severity[sev] = []
            by_severity[sev].append(f)

        report = [
            "# Code Review Report",
            f"**Total Findings**: {len(self.findings)}",
            "",
            "## Summary by Severity",
        ]
        for sev in ["critical", "major", "minor", "info"]:
            count = len(by_severity.get(sev, []))
            report.append(f"- **{sev.capitalize()}**: {count}")
        report.append("")

        for sev in ["critical", "major", "minor", "info"]:
            items = by_severity.get(sev, [])
            if items:
                report.append(f"## {sev.capitalize()} Issues")
                for item in items[:5]:
                    report.append(
                        f"- [{item['type']}] {item['file']}:{item['line']}"
                    )
                    report.append(f"  {item['description']}")
                report.append("")

        return "\n".join(report)

Project 3: Automated Refactoring

An agent that safely restructures code while preserving behavior.

python

class RefactoringAgent:
    def __init__(self, agent):
        self.agent = agent
        self.changes = []

    async def refactor(self, file_path, strategy):
        content = await self.agent.read(file_path)

        if strategy == "extract_function":
            result = await self._extract_function(content, file_path)
        elif strategy == "rename_variable":
            result = await self._rename_variable(content, file_path)
        elif strategy == "add_type_hints":
            result = await self._add_type_hints(content, file_path)
        else:
            return {"status": "error", "message": f"Unknown strategy: {strategy}"}

        return result

    async def _extract_function(self, content, file_path):
        lines = content.split("\n")
        # Find long functions (simplified heuristic)
        changes = []
        i = 0
        while i < len(lines):
            if lines[i].startswith("def ") and i + 30 < len(lines):
                # Function is long, suggest extraction
                func_name = lines[i][4:].split("(")[0]
                changes.append({
                    "type": "extract_function",
                    "function": func_name,
                    "start_line": i + 1,
                    "length": "30+ lines",
                    "suggestion": f"Extract helper functions from {func_name}"
                })
                i += 30
            i += 1

        return {"status": "analyzed", "changes": changes}

    async def _rename_variable(self, content, file_path):
        # Find poorly named variables
        import re
        lines = content.split("\n")
        changes = []
        bad_names = re.findall(r'\b([a-z]{1,2})\s*=', content)
        for name in set(bad_names):
            if name not in ["id", "ok", "no"]:
                changes.append({
                    "type": "rename_variable",
                    "old_name": name,
                    "suggestion": f"Rename '{name}' to a more descriptive name"
                })
        return {"status": "analyzed", "changes": changes}

    async def _add_type_hints(self, content, file_path):
        lines = content.split("\n")
        changes = []
        for i, line in enumerate(lines, 1):
            if line.startswith("def ") and "->" not in line:
                func = line[4:].split("(")[0]
                changes.append({
                    "type": "add_return_type",
                    "function": func,
                    "line": i,
                    "suggestion": f"Add return type hint to {func}"
                })
        return {"status": "analyzed", "changes": changes}

💡Tip

When refactoring with agents, always use a version control system. Create a branch, apply changes incrementally, and run tests after each change. If a refactoring breaks tests, the agent can analyze the failure and self-correct.

Project 4: Test Generation

An agent that creates comprehensive test suites from source code analysis.

yaml

# skills/test-generator/skill.yaml
name: test-generator
description: Generate comprehensive test suites from source code analysis
version: "2.0.0"
instructions:
  - path: instructions/analyze-source.md
  - path: instructions/generate-tests.md
  - path: instructions/validate-tests.md
tools:
  required:
    - read
    - write
    - bash
  optional:
    - grep
resources:
  - path: templates/pytest-test.py.hbs
  - path: templates/jest-test.ts.hbs
autoload:
  enabled: true
  matchPattern: "generate tests|create tests|add test coverage"

python

class TestGenerator:
    def __init__(self, agent):
        self.agent = agent

    async def generate_tests(self, source_path):
        content = await self.agent.read(source_path)
        tree = self._parse_functions(content)
        test_file = self._build_test_file(source_path, tree)
        test_path = source_path.replace("src/", "tests/").replace(".py", "_test.py")
        await self.agent.write(test_path, test_file)
        return {"test_file": test_path, "test_count": len(tree)}

    def _parse_functions(self, content):
        import ast
        tree = ast.parse(content)
        functions = []
        for node in ast.walk(tree):
            if isinstance(node, ast.FunctionDef):
                functions.append({
                    "name": node.name,
                    "args": [a.arg for a in node.args.args],
                    "decorators": [
                        d.id for d in node.decorator_list
                        if isinstance(d, ast.Name)
                    ]
                })
        return functions

    def _build_test_file(self, source_path, functions):
        module_name = source_path.split("/")[-1].replace(".py", "")
        lines = [
            f'"""Tests for {module_name} module."""',
            "import pytest",
            f"from {module_name} import (",
        ]
        for func in functions:
            lines.append(f"    {func['name']},")
        lines.append(")")
        lines.append("")

        for func in functions:
            lines.extend(self._generate_test_case(func))

        return "\n".join(lines)

    def _generate_test_case(self, func):
        name = func["name"]
        params = func["args"]

        test_cases = [
            f"def test_{name}_basic():",
            f'    """Test basic {name} functionality."""',
            f"    # TODO: Implement test for {name}",
            f"    # result = {name}({', '.join(params)})",
            f"    # assert result is not None",
            f"    pass",
            "",
            f"def test_{name}_edge_cases():",
            f'    """Test {name} with edge case inputs."""',
            f"    # TODO: Add edge case tests",
            f"    pass",
            "",
        ]
        return test_cases

Project 5: Documentation Generation

python

class DocGenerator:
    def __init__(self, agent):
        self.agent = agent

    async def generate_docs(self, project_root):
        files = await self.agent.glob(f"{project_root}/**/*.py")
        docs = []

        for file in files[:20]:
            content = await self.agent.read(file)
            module_doc = self._extract_module_doc(content)
            functions = self._parse_functions(content)

            doc_section = [
                f"## Module: {file}",
                "",
                module_doc or "No module documentation.",
                "",
                "### Functions",
                "",
            ]
            for func in functions:
                doc_section.extend([
                    f"#### `{func['name']}({', '.join(func['args'])})`",
                    "",
                    func["docstring"] or "No documentation.",
                    "",
                ])
            docs.append("\n".join(doc_section))

        return "# API Documentation\n\n" + "\n---\n".join(docs)

    def _extract_module_doc(self, content):
        import ast
        try:
            tree = ast.parse(content)
            return ast.get_docstring(tree)
        except SyntaxError:
            return None

    def _parse_functions(self, content):
        import ast
        try:
            tree = ast.parse(content)
            functions = []
            for node in ast.walk(tree):
                if isinstance(node, ast.FunctionDef) and not node.name.startswith("_"):
                    functions.append({
                        "name": node.name,
                        "args": [a.arg for a in node.args.args],
                        "docstring": ast.get_docstring(node)
                    })
            return functions
        except SyntaxError:
            return []

Practice Exercises

Practice Question

What is the recommended approach when an agent refactoring breaks existing tests?

Practice Question

In the code review project, what security issue does the agent check for?

Practice Question

What tool does the test generator use to parse source code and identify functions?

Practice Question

What is the purpose of using a branch for agent-driven refactoring?

Practice Question

In the blog API generation project, what framework is used for the generated API?

Practice Question

What file naming convention does the test generator use when creating test files?

Practice Question

What information does the documentation generator extract from source code?

Practice Question

Why does the code review agent use 'deniedTools: [write, edit]'?

✅**Key Takeaways**

Real-world agent projects span code generation, review, refactoring, testing, and documentation
Code generation agents create complete modules from specifications with models, routes, and tests
Code review agents should be read-only, inspecting code for bugs, security issues, and style problems
Refactoring agents should use version control branches for safe, reversible changes
Test generation agents create comprehensive test suites by parsing source code structure
Documentation agents extract module and function documentation from source code automatically
Each project combines multiple skills: tool use, planning, context management, and safety patterns
Real agents handle errors, adapt plans, and learn from intermediate results
The capstone demonstrates the full power of agentic AI: autonomous, multi-step software engineering

Progreso100%

Best Practices & Patterns

Volver a Cursos

Contenido del Curso

1. Introduction to AI Agents 2. Agent Architecture 3. Skills & Capabilities 4. Context Management 5. Agent Pipelines 6. OpenCode as Development Tool 7. Building Custom Agent Skills 8. Prompt Engineering for Agents 9. Best Practices & Patterns 10. Real-World Agent Projects