Hypothesis-driven debug mode for Claude Code, Codex, and other agent platforms. Instead of guessing fixes, the agent generates root cause hypotheses, instruments your code with targeted runtime logs, asks you to reproduce via structured choices, analyzes the evidence, applies a targeted fix, then cleans up all instrumentation automatically.
Ten Claude Code skills for writing an ML paper, abstract to rebuttal. Recipes, axioms, and mechanical checklists reverse-engineered from papers that got in.
A local dashboard over your own Claude Code usage: tokens, cost, prompts, sessions, tool calls and failures — parsed from the JSONL transcripts Claude Code already writes, with an optional live feed driven by HTTP hooks.
Spec-driven development skills for AI coding agents: 10 human-gated steps from idea to docs. Claude Code plugin + SKILL.md for Codex, Cursor, Gemini CLI, OpenCode
Data-driven Dota 2 tier list installed as an in-game hero grid. Agent skill for Claude Code & Codex (or plain Python): the best heroes per role on the current patch, filtered for your MMR bracket. Built to climb MMR.