code-agent

Automating a multi-file bug fix requires autonomous workspace exploration, scoped phased implementation, and iterative test verification. The code-agent reads project files to understand architecture, proposes a plan, implements changes across modules, and runs the test suite until correctness criteria are met.

Autonomous multi-module refactoring works best when split into explore, implement, integrate, and verify phases with checkpoints. This approach reads project dependencies and patterns before making changes, ensuring the refactor aligns with existing architecture while running automated tests to confirm correctness.

Automating test suite repair and CI test triage is possible by autonomously investigating hanging or failing tests across multiple modules. The agent identifies the root cause, proposes a scoped fix, implements the changes, and iteratively runs verification steps until the failing test suite passes.

Autonomous code investigation supports feature additions within a project's workspace by first reading project files to understand dependencies and patterns. It then proposes a phased implementation plan, applies the feature changes, and executes automated verification steps to ensure integration succeeds.

Automated code verification handles iterative test execution by running tests, reporting results, and iterating through changes until correctness criteria are met. This ensures that multi-file bug fixes, refactors, and feature additions are validated within the workspace before completion.

You should avoid using an autonomous agent for repository tasks when low-level manual investigation is preferred over intent-based decisions, or when a project's workspace lacks automated tests. The agent relies on running test suites to iteratively verify correctness, which is essential for its phased implementation approach.