From devcore
Guides systematic debugging of hard bugs via rubber ducking, code flow tracing, hypothesis testing, targeted logging, and agent investigations. Use when bugs resist quick fixes.
npx claudepluginhub captaincrouton89/crouton-kitThis skill uses the workspace's default tool permissions.
Quote code snippets and explain what each chunk *should* do vs *actually* does. Don't assume—verbalize the logic. Discrepancies reveal bugs.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Checks Next.js compilation errors using a running Turbopack dev server after code edits. Fixes actionable issues before reporting complete. Replaces `next build`.
Quote code snippets and explain what each chunk should do vs actually does. Don't assume—verbalize the logic. Discrepancies reveal bugs.
Trace data from entry point to failure. At each transformation: what goes in, what comes out? Mark where expectations diverge.
Failure-prone boundaries: async, serialization, type coercion, null propagation, state mutations.
Don't change code hoping it helps—that creates noise.
Log at decision points and async boundaries, not everywhere.
Workflow:
For complex/unfamiliar code sections acting as a blackbox:
When stuck: Spawn 2-3 senior-advisor agents in parallel with different perspectives (pragmatist, architect, skeptic).
Avoid biasing agents: Pass them relevant file paths and the observed behavior, but not your hypotheses or assumptions. Let them form independent conclusions.
If no, keep investigating.