Guide · Agent skill
How to use the Design Action Combat agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Design Action Combat is an agent skill that teaches an AI agent to build action combat for web games as explicit, testable state machines instead of animation-driven guesswork. This guide explains what the skill changes about the agent's behaviour, how to install it, how to ask for help in plain language, and what to watch for. By the end you will be able to add the skill to your project and request timing windows, hit resolution, boss phases or combat tests with clear expectations.
Design Action CombatMengTo/Skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the design-action-combat skill as a ZIP from this page.
- 2Unzip the skill. Unzip the archive so you have the skill folder containing agents/openai.yaml and SKILL.md.
- 3Place the folder. Move it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside your project.
- 4Or upload in the apps. Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.
- 5Describe your combat task. Ask in plain language about attack timing, guard windows, boss phases or combat tests.
- 6Provide project context. Share your frame rate, current action values, and where collision events are produced.
- 7Verify in play. Run the suggested deterministic tests and check the feel at normal camera distance.
What does the Design Action Combat skill make the agent do?
Once installed, the agent approaches combat as a set of states with visible timing. For each player or enemy action it defines startup, active window, recovery, cancellation rules, resource cost, contact shape, damage or posture outcome, cooldown and feedback. It also aims to make telegraph, danger, contact and recovery readable at normal camera distance.
- Hits, blocks, parries and interrupts resolve from authoritative collision or contact events, not from animation frames.
- Each contact is applied once, using stable action and target identifiers.
- Defenses are given distinct purposes such as spacing, dodge, timed guard, interruption or resource trade.
- Boss phases change decision pressure without invalidating timing the player has already learned.
- Combat tests cover early and late inputs, wrong direction, out-of-range contact, interrupted actions, cooldown boundaries and repeated inputs.
Install the skill from the ZIP download
Download the skill as a ZIP from this page, then unzip it. The skill folder contains agents/openai.yaml and SKILL.md. Put the unzipped skill folder in one of these locations:
- Personal skills: ~/.claude/skills/ so it is available across your projects.
- Project skills: .claude/skills/ inside your game project, so it applies only to that project.
- Claude apps: upload the skill under Settings → Capabilities → Skills.
Trigger the skill with natural requests
You do not need a special command. Describe the combat problem in ordinary language and mention the kind of work you want. The skill's description covers attack timing, guard and dodge windows, hit contact, posture, lock-on, weapons, boss phases, combat feedback and deterministic combat tests, so requests in those areas are the ones it is written for.
- "Set up startup, active and recovery frames for a sword light attack in my web game."
- "Design a timed guard window for the shield enemy and explain how it should differ from a dodge."
- "Make the hit resolution use collision events and stop one swing from hitting the same enemy twice."
- "Plan deterministic tests for early and late dodge inputs and for interrupted attacks."
- "Suggest three phases for the boss and how each one raises pressure without making its patterns unlearnable."
Give the agent what it needs for good results
The source lists no required tools or packages. The most useful input is your own context: the game's frame rate, the current attack and defense values, how collisions are detected in your code, and the test framework you use. Paste or point the agent at the relevant files so its timing proposals match your engine and state logic.
- Your target frame rate and typical camera distance.
- Existing action data, such as attack durations and cooldowns.
- Where collision or contact events are produced in your code.
- Any reduced-motion or input-alternative requirements you already have.
Safety notes: scripts and network
According to the item details, this skill does not run scripts and does not use the network. Its files are agents/openai.yaml and SKILL.md. The skill is guidance for the agent; it does not generate finished art, animation or sound assets, so you still need to produce those yourself. Review any code the agent writes before merging it, as you would with any generated code.
- Runs scripts: no.
- Uses network: no.
- Licence: MIT.
Tips for better combat design output
Ask for one verb at a time. A complete timing table for a single attack is easier to check than a whole moveset. Ask for the reasoning behind a number, not just the number, so you can judge whether it suits your game. Finally, ask the agent to list the edge-case tests it expects, then run them against your build.
- Start with one attack or one enemy, then expand.
- Ask why a defense exists: spacing, dodge, timed guard, interruption or resource trade.
- Check that the player's intent, contact feedback, health or posture changes and target state all agree during play.
- Test at realistic frame rate and camera distance, not only in a debug view.
What people use it for
- Defining startup, active, recovery and cooldown values for a melee attack.
- Designing a timed guard or dodge window that rewards precise input.
- Making hit and block outcomes resolve from collision events with no double hits.
- Planning boss phases that increase pressure while keeping patterns learnable.
- Writing deterministic tests for early or late inputs, interrupted actions and cooldown edges.
Frequently asked questions
How do I install the design-action-combat skill?
Download the ZIP from this page and unzip it. Place the skill folder in ~/.claude/skills/ for personal use or in .claude/skills/ inside a project. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
Does the design-action-combat skill need anything else installed?
The item lists no requirements. It runs no scripts and uses no network. Your own project context, such as frame rate, action values and your test framework, will make its output more useful.
Is the design-action-combat skill free to use?
The skill is released under the MIT licence, as stated on this page. Check the licence file in the source repository for the full terms.
What kinds of games is this skill for?
It targets action combat in web games. Its guidance may not map cleanly to non-game projects or to game engines outside web contexts. It also does not generate art, animation or sound assets.
How is this different from just asking an AI for combat ideas?
The skill pushes the agent toward explicit state machines, authoritative collision events and deterministic tests rather than animation-driven guesses. That gives you timing you can check and test, though you should still review the output yourself.
Why isn't my agent using the skill?
Check that the unzipped folder sits directly in ~/.claude/skills/ or .claude/skills/, or that it was uploaded under Settings → Capabilities → Skills. Then describe a combat task that matches the skill's areas, such as attack timing, guard windows or boss phases.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an indie or hobbyist developer building action combat in a web game, a designer who wants clear timing and balance reasoning for enemies and bosses, or a developer writing automated combat tests. Pick something else if you work outside web games, use a game engine outside web contexts, or need finished art, animation or sound.
Download the skill