2.0 Inside a Coding Agent: Overview of Skills, Tools, Context | Asha

Why can the same AI model behave differently in two coding workspaces? Asha separates the model, harness, context, tools, project guidance, skills, delegated specialists, and enforced permissions. Follow concrete prompt and configuration examples across Codex, Claude Code, and Cursor, then connect the workspace to MCP, worktrees, sandboxes, and recovery. This standalone overview is chapter 00 of lesson 2 in AI-Native Software Engineering. The following chapters provide deeper worked examples. CHAPTERS 00:00 Open the agent's workspace 00:20 Model, harness, context, tools, and controls 01:21 Task prompts and project guidance 02:01 Repository evidence and trust 02:31 Build a reusable skill 03:32 Define and delegate to a specialist 04:53 MCP and service authorization 05:34 Worktrees, sandboxes, and durable state 06:24 Codex configuration examples 06:55 Claude Code configuration examples 07:25 Cursor configuration examples 07:55 Portable intent and a bounded review prompt 08:26 Choose the right mechanism 08:46 Build a small, inspectable workspace COURSE AND SOURCES https://storage.googleapis.com/intera... https://learn.chatgpt.com/docs/custom... https://learn.chatgpt.com/docs/build-... https://learn.chatgpt.com/docs/agent-... https://code.claude.com/docs/en/memory https://code.claude.com/docs/en/skills https://code.claude.com/docs/en/sub-a... https://cursor.com/docs/rules https://cursor.com/docs/skills https://cursor.com/docs/subagents https://modelcontextprotocol.io/docs/... The maintenance contract and configuration snippets are teaching examples. Product behavior and supported configuration can change; verify the installed version and official documentation. Created with a fictional AI-generated presenter and illustrative cinematic scenes. Exact diagrams and code are authored educational graphics. Demonstration only — verify with the source. #AINativeEngineering #CodingAgents #SoftwareEngineering