Workflow General Debugging
Testing & Qualityworkflow-general-debuggingWorkflow General Debugging is an Agent Skill for testing & quality published in the v8/v8 repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Workflow for general debugging tasks without the overhead of issue tracking. Use when investigating crashes, hangs, or unexpected behavior where no specific issue ID is provided. Do not use when issue ID is provided.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/v8/v8/blob/HEAD/agents/skills/workflow-general-debugging/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/workflow-general-debugging, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Skill: Workflow - General Debugging
Workflow for general debugging tasks in a pristine branch or project work, without the overhead of issue tracking. Use this skill when investigating crashes, hangs, or unexpected behavior where no specific issue ID is provided.
Principles
- Lightweight & Focused: Focus on the immediate problem without the
overhead of issue tracking conventions (e.g.,
regress-<id>naming,Bug:tags) unless requested. - Efficient Investigation: Prioritize quick root cause analysis using GDB and targeted code reading over exhaustive documentation.
- Pristine Branch Work: Assume you are working on a clean branch or ongoing project work where changes are local.
- Architectural Skepticism: Always question your fix. Ensure it addresses the root cause and not just a symptom.
Workflow
1. Problem Triage
- Reproduce the issue if possible.
- Analyze the symptom (crash, hang, incorrect output).
- If a repro script is available, do not modify it directly. Create a scratch copy for investigation.
2. Static & Dynamic Analysis
- Use
code_searchorgrep_searchto locate relevant code based on stack traces or symptoms. - Use GDB (
gdb-mcp) to inspect runtime state, set breakpoints, and trace execution. - Focus on understanding the immediate failure context before expanding the scope.
3. Propose Fix
- Develop a minimal fix targeting the root cause.
- Follow V8 Best Practices.
- Architectural Skepticism: Argue against your own proposal. Does it disable optimizations? Does it violate invariants?
4. Verify
- Create a Regression Test: Design and implement a clean, semantic regression test. Verify it fails on the unfixed code and passes cleanly on the fixed code.
- Verify the fix against the repro or by running relevant unit tests.
- Ensure no new failures are introduced in related areas.
5. Finalization
- Format the code using
git cl format. - Present the Fix & Test: Present both the fix diff and the accompanying working regression test to the user/main agent.
- Do not create commits or upload CLs unless explicitly requested.