Skills

game-sensitivity-coach: Carry Your Familiar Sensitivity to a New Game and Tune It with Gameplay Evidence

If you want to carry a sensitivity you're used to from one game into another, or tune it based on evidence from real gameplay recordings, this agent skill is the right fit. Before use, you need Python 3.8+, FFmpeg, and a Windows gaming environment, and conversion suggestions won't automatically change your settings.

Official source · game-sensitivity-coach

Install

git clone https://github.com/DenisSergeevitch/game-sensitivity-coach.git ~/.agents/skills/game-sensitivity-coach

Who It's For

  • FPS players who want to carry a mouse sensitivity they're familiar with from one game into a new one, or do evidence-based sensitivity tuning using real gameplay recordings.
  • Anyone with a local agent on their computer that can run commands, read files, and view images (official documentation uses Codex, Claude Code, and other Agent Skills hosts as examples), and is willing to place the skill folder using the Codex / Claude Code directory structure.
  • Before use, confirm your environment meets the officially listed requirements: Python 3.8+ (standard library only, no Python packages or AI API key needed), FFmpeg (for recording video and exporting evidence; must be in PATH or specified via --ffmpeg), and a Windows gaming PC (for the bundled live recording).
  • Not suitable when: no explicit evidence available; for example, it's unconfirmed whether non-Windows gaming environments are supported, whether there are anti-cheat restrictions, or whether your specific game is covered — please verify in the repository before use.

First Try in 3 Minutes

Prerequisites: Python 3.8+ and FFmpeg ready, and an agent that can execute local commands and read files.

  1. Install into the agent's skill directory. The official docs also provide a method of downloading game-sensitivity-coach.zip from GitHub Releases and extracting it (with a SHA-256 checksum); when extracting, the entire game-sensitivity-coach folder (including scripts, references, assets) must be kept together. Cloning directly is equivalent:

    git clone https://github.com/DenisSergeevitch/game-sensitivity-coach.git ~/.agents/skills/game-sensitivity-coach

    For Claude Code, the corresponding location is ~/.claude/skills/game-sensitivity-coach/ (or project-level .claude/skills/game-sensitivity-coach/); for other Agent Skills hosts, copy it to their documented skill directory.

  2. For your first task, pick a low-risk, reversible usage: describe your current mouse DPI/sensitivity and the target game to the agent, and ask it to do a cross-game sensitivity conversion. This kind of calculation only produces suggestion text.

  3. Success signal: the agent gives a conversion result with reasoning based on your input parameters. The official docs explicitly state that suggestions alone won't change your settings — modifications happen only when you explicitly ask to "apply" — so the first attempt has no side effects.

Use Cases

Case 1: Cross-game conversion. You're switching from game A to game B; give the agent both game names plus your current sensitivity/DPI and ask it to convert. Per the official description, the skill performs the conversion and provides usable values to try; this process involves only computation and text output.

Case 2: Gameplay review. Ask the agent to record a limited gameplay clip (the skill handles recording, checking video frames and mouse movements; live recording is provided on Windows), then ask it to distinguish whether an issue is sensitivity-related or due to timing, recoil control, or target movement, and to give settings recommendations to keep or try. The result should include evidence based on the actual footage and one clear recommendation; whether to apply it remains your decision.

Usage Boundaries

  • Files and commands: the skill needs the agent to run local commands, read files, and inspect images; recording and exporting evidence rely on FFmpeg. Whether it reads or writes files beyond your game configuration is not stated in the official evidence.
  • Network access: no explicit evidence; unknown.
  • Side effects and confirmation points: the official docs explicitly state that "recommendations alone don't change settings; an explicit application request is required"; confirmation mechanisms for other actions are not described in the evidence.
  • Runtime environment: Windows (live recording on the gaming PC); whether Linux/macOS works is not stated by the official docs.
  • Clients: the official docs list directory mappings for Codex and Claude Code; for other Agent Skills hosts, place it in their respective skill directories. No other installation methods (such as a plugin marketplace) are confirmed.
  • Not yet confirmed: the list of supported games, supported languages/text types, whether other named workflows or check criteria exist, and details of security boundaries. Please consult the repository README and the original skill files before use.