Course Outline
From Autocomplete to Agents: Understanding Agent Failures
• Components of a coding agent: model, harness, tool surface, context, and permissions
• Positioning each tool: Claude Code, GitHub Copilot, Cursor, Codex CLI, and Gemini CLI
• A taxonomy of failure: incorrect context, unsuitable tools, lack of feedback, and unbounded autonomy
Demonstration: The same task executed successfully versus poorly, side by side
Context Engineering
• Treating the context window as a budget: determining what earns a spot within it
• AGENTS.md, CLAUDE.md, .cursor/rules, copilot-instructions.md — one concept, varied filenames, single source of truth
• Conventions, build and test commands, and architectural boundaries
• Retrieval versus explicit context; task decomposition and sub-agents
Lab: Draft repository context for an unfamiliar Python service, then re-run a failing task to compare outputs
Reusable Workflows and Agent Skills
• Selecting the right abstraction: instruction file, skill, custom command, or plain script
• Structure of a skill: triggers, instructions, bundled scripts, and progressive disclosure
• Portability across tools, and identifying where lock-in begins
• Versioning, review processes, and distribution within a team; common anti-patterns
Lab: Create and test a reusable workflow that enforces house coding standards
MCP: Connecting Agents to Real Systems
• Architecture: clients, servers, tools, resources, and prompts; stdio and HTTP transports
• Servers that justify their presence: Git hosting, issue trackers, databases, browsers, internal APIs
• When a CLI or script outperforms an MCP server
• Tool-surface hygiene: why excessive tools reduce reliability
Lab: Configure MCP servers and handle a ticket end-to-end — issue, branch, patch, tests, pull request
Feedback Loops and Evaluation
• Tests, types, and linters as the agent’s ground truth; test-first approaches as a control mechanism
• CI as the outer loop, and review discipline for agent-authored diffs
• Golden-task evaluation sets: what to measure and how to identify regressions
• Cost and latency as primary metrics
Lab: Construct a small evaluation set and score two agent configurations against it
Security and Guardrails
• Prompt injection via issues, pull requests, READMEs, dependencies, and fetched pages
• Permission models: allowlists, approvals, read-only tools, network egress control
• Secret hygiene and sandboxing: containers, ephemeral credentials, limiting blast radius
• Supply-chain risks in third-party MCP servers and shared skills
Lab: Observe an agent being hijacked by a poisoned repository, then harden the setup to prevent recurrence
Rolling This Out to a Team
• A staged adoption path; what to standardize and what to leave to individual choice
• Metrics that indicate genuine value versus those that do not
Requirements
• Proficient understanding of Python, Git, and the command line
• Prior exposure to an AI coding assistant
• NobleProg will configure Dadesktop VMs for participants containing Docker, VS Code, and Python 3.11 or later
• A functional AI coding assistant of your choice: Claude Code, GitHub Copilot, Cursor, Codex CLI, or Gemini CLI. Labs are tool-agnostic, with instructions provided for each
Audience
• Software engineers, tech leads, and architects utilizing AI coding assistants without achieving reliable outcomes
• Platform and developer-experience engineers deploying AI tools across teams
• Engineering managers establishing standards, guardrails, and success metrics
Testimonials (3)
The trainer is patient and very helpful. He knows the topic well.
CLIFFORD TABARES - Universal Leaf Philippines, Inc.
Course - Agentic AI for Business Automation: Use Cases & Integration
Good mixvof knowledge and practice
Ion Mironescu - Facultatea S.A.I.A.P.M.
Course - Agentic AI for Enterprise Applications
The mix of theory and practice and of high level and low level perspectives
