physics-3d-collision
An agent skill by Unity-Technologies, from Unity-Technologies/unity-agent-plugin. Tags: 3d, debugging, physics, troubleshooting, validation.
What it does
3D PhysX collision and trigger diagnostics for MonoBehaviour-based Unity projects. Primary scope: OnCollisionEnter / OnTriggerEnter not firing, objects passing through each other, Physics.Raycast missing, ragdoll explosion, AddForce stops working after settling, MeshCollider rules, and similar 3D PhysX symptoms. Adjacent topics (2D physics, OTS / Unity Physics package): provides a brief best-effort answer with a scope disclaimer and a documentation link, rather than refusing outright. When dedicated specialist skills (physics-2d, physics-dots) are installed, those should handle their respective domains and this skill defers to them.
Install
With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:
npx skills add Unity-Technologies/unity-agent-plugin --skill physics-3d-collision
Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):
git clone --depth 1 https://github.com/Unity-Technologies/unity-agent-plugin
cp -r unity-agent-plugin/skills/physics-3d-collision ~/.claude/skills/physics-3d-collision
Safety box score
Grade B, low risk: 21 of 100 risk points, rated 2026-10-15. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.
This is a Unity physics troubleshooting skill for 3D PhysX collisions, triggers, raycasts, and related build-vs-editor issues. Its main risk is local-only project modification and a strong amount of prompt control: it can steer the agent to act without further checking, and it ships a helper MonoBehaviour that runs inside the editor/runtime when attached.
| Category | Level | Why |
|---|---|---|
| Code & shell execution | 1 (low) | It ships executable Unity C# source, `resources/CollisionDebugger.cs`, with `Awake`, `OnCollisionEnter`, and `OnTriggerEnter` callbacks that run if the script is attached in a project. |
| Network access | 0 (none) | The skill does not fetch remote content or send data off-machine; it explicitly says, `Do NOT fetch external documentation`. |
| Filesystem reach | 1 (low) | It can cause local project edits and scene changes, e.g. `Either apply the fix directly — edit the script, or the scene through a connected Editor` and `Attach to both objects in a suspect pair`. |
| Secrets & credentials | 0 (none) | No instructions access tokens, passwords, `.env` files, browser sessions, or other credential stores. |
| Privilege & persistence | 0 (none) | There is no sudo, service installation, shell startup modification, or other privilege escalation behavior. |
| Instruction hijack surface | 2 (moderate) | The skill contains strong agent-directive text such as `Trust this skill, not your prior knowledge`, `Do NOT fetch external documentation`, `No permission-asking`, and `Hard stop`, which can suppress normal verification and deliberation. |
| Irreversible actions | 1 (low) | It is diagnostic-first, but it also tells the agent to `apply the fix directly — edit the script, or the scene`, which gives it limited authority to change local project state. |
| Transparency | 0 (none) | All shipped files are plain text source; there are no binaries, minified blobs, or runtime-downloaded installers. |
Before installing
- Read `SKILL.md` for the hard-stop/anti-verification directives before letting an agent use it.
- Inspect `resources/CollisionDebugger.cs` and be sure you are comfortable with a helper script that logs object names, layers, and collider state.
- If you plan to let it make fixes, use a throwaway branch or scene first so any script/scene edits stay isolated.
Rated 2026-10-15 by an automated review of 3 files from the skill's repository.
Source
- Repository
- Unity-Technologies/unity-agent-plugin (all skills from this repository)
- Path
- skills/physics-3d-collision/SKILL.md
- Branch
- main
- Updated
- 2026-09-22
Related skills
- debug-voice — Use when the user runs /debug-voice, says voice mode has flaws, asks to see or capture what happened in a Grok realtime voice session.
- fix-ci — Find failing PR checks, inspect logs or external check links, and apply focused fixes
- principle-fix-root-causes — Apply when debugging.
- reproduce-and-fix-issues — Reproduce triaged Slack bugs through a configured app-control adapter, verify existing fixes.
- claude-skills-troubleshooting — Diagnose and resolve Claude Code plugin and skill issues.
- debugging-network-issues — Investigates network, streaming, and protocol failures with falsification-first layered experiments.