This skill turns a coding agent such as Claude Code or Codex into a development partner for fiction, screenplays, outlines, scenes and characters. You give it real story material, and it looks for the single problem that matters most, sets up competing explanations for it, and tries to disprove its own diagnosis with counterexamples and what-if tests before suggesting the smallest change that would fix it. It works in loops, three by default, re-checking the story after each change and stopping when further passes stop helping or when the best move is to leave the story alone. A separate mode handles producer-style review of full-season or multi-episode outlines. You get a tested diagnosis and a targeted revision plan instead of a blanket rewrite.
Write story chapters from an approved outlineThis skill writes fiction chapters in an outline-first order. It gathers context for the chapter with a story command-line tool that keeps later chapters out and stays within a token budget, builds a scene-by-scene outline for your approval, and only then writes the full prose. Afterward it records scene dates, continuity state and setups planted or paid off, and rebuilds the project's registries. It expects a story bible plus a character file or more, and ideally a plot structure, and it passes line editing to a companion skill.General assistantFictionWriting
Reason through a hard question as a panel of three expertsThis skill turns a question into a discussion among three simulated experts, shown as one markdown table. Each expert explains their reasoning, builds on the others, admits mistakes and revises, until the panel reaches one answer. It is a single-prompt form of Tree of Thoughts for puzzles, math, logic, estimates and strategy questions; the experts are model-generated roles, not real specialists.General assistant
Cut filler and repetitive patterns from proseStephen Turner's deslop skill edits prose, including scientific writing, to cut repetitive sentence structures, filler, vague claims and flat rhythm. Its checks are style edits, not a test of who wrote a text.General assistant
Run an OODA loop on an unclear decisionneurofoo's ooda skill runs an OODA (Observe, Orient, Decide, Act) loop: gather facts, name causes, list options and produce a verified implementation plan for unclear or complex decisions.General assistantoodadecisions
Write optimized prompts for a named AI toolThis skill writes one ready-to-paste prompt for a named AI tool, such as an LLM, Cursor, Midjourney, or a video generator. It confirms the target tool when that is unclear, pulls the task, output format, constraints, and audience out of a rough idea, and asks at most three clarifying questions. It prefers simple techniques over those that raise the risk of invented content, and never asks a model to reveal hidden reasoning. The result is a single prompt block with a one-line note on what was optimized.General assistantPrompting
Make animated GIFs for SlackThis skill makes animated GIFs that fit Slack, at emoji or message size, with frame rate, color, and length limits. It draws frames from shapes or a provided image, saves an optimized file, and checks it against Slack's limits.General assistantSlackImages
