llm-enhanced-test

agents contains all agents used to generate WAMR test cases with LLMs. If you modify existing agents or add new ones, keep the current directory structure.

How to use these agents

The agents directory includes two parts:

1. Agents for generating new test cases.
2. Agents for reviewing and fixing generated test cases (tests-pipeline.md, tests-review.md, tests-verify.md, tests-fix.md).

When using these agents, copy AGENTS.md, CLAUDE.md, .claude, and .kode to the WAMR repository root so the agents can better understand the project structure.

Copy scripts under scripts into the tests/unit directory, which is the agents' working directory.

For example, to run the tests-pipeline agent with Claude Code CLI:

claude --dangerously-skip-permissions < path/to/agents/.claude/agents/tests-pipeline.md

In Claude, using --dangerously-skip-permissions bypasses all permission checks. Make sure you run it only in a secure environment.