Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
69 changes: 0 additions & 69 deletions skills/levelplay-unity-integration/CHANGELOG.md

This file was deleted.

78 changes: 0 additions & 78 deletions skills/levelplay-unity-integration/README.md

This file was deleted.

20 changes: 20 additions & 0 deletions skills/levelplay-unity-integration/SECURITY.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
# Security notes: levelplay-unity-integration

When the user chooses it, this skill has the agent call the LevelPlay Management API to find or create the user's LevelPlay app and ad units (`references/management-api.md`). That means authenticated network calls with the user's account credentials, and resources created on the user's account. It is intentional, and it is limited by the safeguards below.

## Accepted risks

| Risk | Capability | Why it is accepted |
|---|---|---|
| Authenticated network calls | Sends HTTPS requests to `platform.ironsrc.com` with a secret key and refresh token read from the environment variables `LEVELPLAY_SECRET_KEY` and `LEVELPLAY_REFRESH_TOKEN` | The user sets the variables themselves and chooses this path over the dashboard. The calls go only to the LevelPlay platform host that issued the credentials, and only to the documented auth, applications and ad units endpoints. |
| Resources that cannot be deleted through the API | Creates apps and ad units on the user's LevelPlay account | The API has no delete call, so the agent lists what exists first, reuses it, and creates only what is missing, after the user confirms a table of exactly what will be created. |

## Mitigations

- **Credentials from the environment only.** The agent never asks the user to paste the secret key or refresh token, references them only as `$LEVELPLAY_SECRET_KEY` and `$LEVELPLAY_REFRESH_TOKEN` in commands, and checks them with `test -n` rather than printing them.
- **Never echoed, logged or written.** The credentials and the bearer token are never printed, never sent with verbose or tracing flags, and never written to a project file, an `.env` inside the project, a commit or generated C#.
- **Explicit confirmation before any create call.** The agent shows one table of every app and ad unit it will create, states that they cannot be deleted through the API, and proceeds only on an explicit yes.
- **Reuse before create.** Existing apps are matched by name and platform, and existing ad units by format, so a repeated run creates nothing new.
- **Project names never enter a shell command.** The app and ad unit names come from the project's files. The agent writes each create call's JSON body to a temporary file outside the project with its file-writing tool, checks it with `python3 -m json.tool`, sends it with `curl --data-binary @file`, and deletes the file, so a quote or other shell character in a name cannot end the command early or run anything.
- **Documented endpoints only.** The agent calls only the auth, applications and ad units endpoints described in the LevelPlay API documentation. It does not enable networks, change billing or take an app live.
- **No third-party hosts.** No request goes anywhere but `platform.ironsrc.com`, and no response is executed.
Loading
Loading