AI coding assistant
A model-driven tool that writes, edits and runs code in your project, from autocomplete to agents that make whole changes.
In more detail
Run coding agents in a sandbox with project-only writes, allowed network domains and no production credentials. Gate new dependencies, scan diffs for secrets, review small diffs yourself and let CI, not the agent, decide whether the tests pass.
Related terms
- Claude Code — Anthropic's agentic coding tool, which reads a codebase, edits files and runs commands from the terminal or an IDE.
- AGENTS.md — An open format for a project file that tells coding agents the house rules: how to build, test and style code in that repository.
- package hallucination — A coding model recommending a dependency that does not exist, which an attacker can then register with malicious code.
- sandbox — A locked-down environment in which a program runs with very few permissions, so that damage from a bug or attack stays contained.
Taught in
- Designing Tools A Model Can Use — AI Engineering: LLM Apps, RAG & Agents
- Agent SDKs And Human Approval — AI Engineering: LLM Apps, RAG & Agents
- Images, Audio And Speech — AI Engineering: LLM Apps, RAG & Agents
- Working With AI Coding Assistants — AI Engineering: LLM Apps, RAG & Agents
References
- Claude Code docs: Configure the sandboxed Bash tool (filesystem and network isolation)
- Spracklen et al., We Have a Package for You! A Comprehensive Analysis of Package Hallucinations by Code Generating LLMs (USENIX Security 2025)