What problem does it solve? Practicing LeetCode across multiple languages often leaves run results, mistakes, and notes scattered and inconsistent. This Skill structures the whole loop: pick a problem, run the configured command, review the attempt with minimal hints, and merge the learning into one SecondBrain note per problem. ## Core Features & Use Cases - Run-then-review workflow: Executes configured per-language run commands (C++, Java, Kotlin), captures exit codes, stdout/stderr summaries, and classifies failures as compile errors, runtime errors, wrong answers, or timeouts. - Minimal-hint coaching: Reviews code against a rubric with verdicts like accepted, needs_revision, or cannot_prove_correctness, giving direction without full solutions unless requested. - Note patch generation: Produces note patches that merge multi-language attempts into a single problem note while preserving existing front matter, TODOs, and Obsidian links, applying changes only with explicit user confirmation. - Durable mistake memory: Records reusable mistakes, language pitfalls, and spaced-review cues as JSONL under the runtime state directory. - Use Case: After writing a C++ solution for 3Sum, ask the coach to run the configured test command; it captures the wrong-answer failure, identifies duplicate-handling as the root cause, suggests a minimal fix, and drafts a note patch for your SecondBrain. ## Quick Start Ask the coach to run and review my current LeetCode attempt and generate a note patch for the problem.
