SkillsTech Certified AI-Era Software Engineer Associate

The AI-era engineering mindset and human accountability, assistant vs agent, the agentic loop (plan, tools, approval, tests, PR), permission modes, and when to stop or restart the agent.

• Apply the AI-era mindset: AI generates code, humans own the consequences
• Distinguish an AI assistant from an autonomous agent and pick the right one
• Run the agentic loop: plan, tool use, human approval, tests, branch, PR
• Choose permission modes and approval gates for an agent
• Recognize when to stop an agent and when to restart with better context
• Identify categories of AI coding tools (chat, inline, agentic CLI, cloud agent, IDE agent)

Prompt engineering (how you ask), context engineering (what the agent can see), and agent engineering (tools, memory, permissions). Reviewable outputs and injection awareness.

• Frame tasks with role, success criteria, and constraints
• Engineer context: repo instruction files (AGENTS.md / CLAUDE.md), examples, tool rules
• Ask for plans, diffs, tests, risks, and rollback before accepting code
• Recognize prompt injection and context poisoning
• Make agent outputs small, reviewable, and verifiable

Retrieval-augmented generation (embeddings, chunking, retrieval, reranking, grounding, evaluation) and AI application architecture (tool calling, structured outputs, human-in-the-loop, cost/rate limits, fallback).

• Explain the RAG pipeline: embeddings, chunking, indexing, retrieval, reranking
• Use grounding and citations to reduce hallucination
• Apply hybrid search, metadata filters, freshness, and access control
• Decide RAG vs fine-tuning and when NOT to use RAG
• Design AI apps: tool/function calling, structured outputs, human-in-the-loop
• Add cost control, rate limits, caching, and model fallback

The verification discipline for AI-assisted changes: tests, type checks, static analysis, security and dependency scanning, blast-radius review, rollback, observability, and secrets handling.

• Choose the right tests: unit, integration, regression
• Apply type checks, static analysis, and security scanning
• Assess dependency, supply-chain, and license risk
• Review blast radius and architectural impact of a change
• Plan rollback and observability for a risky change
• Handle secrets, environment, and data-migration safety

Owning the repository: key files, architecture, dependencies, change impact, behavior prediction, freshness/staleness, and recognizing overconfidence about AI-generated code.

• Identify key files, architecture, and workflows in a repo
• Explain change impact and risky areas
• Understand dependencies and predict behavior
• Reason about freshness/staleness and what changed since last proof
• Recognize overconfidence in AI-generated code you did not verify

Explaining and defending an AI-assisted change: what the AI did, what you accepted or rejected, what was verified, what risks remain, and how to explain it to a manager and a security reviewer.

• Explain what the AI did vs what you accepted or rejected
• State what was verified and what risks remain
• Explain a change to a manager and to a security reviewer
• Produce a rollback and monitoring note

Security and governance for AI-assisted work: prompt/tool injection, data exfiltration and secret leakage, over-permissioned agents, dependency hallucination, audit trails, and secure MCP usage.

• Mitigate prompt injection, tool injection, and data exfiltration
• Avoid secret leakage, over-permissioned agents, and unsafe shell commands
• Handle untrusted-repo risk and dependency hallucination (fake packages)
• Apply human approval, audit trails, secure MCP usage, and governance