What goes in a Secret asset
Write oneKEY=value assignment per line, in the same format as a .env file:
Use a secret in a chat
1
Turn on the sandbox
The sandbox is what runs code, and it is the only thing that reads secrets. Turn on Sandbox (execute) in the chat’s agent settings. If the sandbox is off, Athena attaches the secret but warns that its values will not be injected.
2
Attach the secret
@-mention the Secret asset in your message, or add it under Sandbox Secrets in the agent settings.
3
Read the keys from code
Every command the agent runs in the sandbox sees each key as an environment variable, and so does a Python script started from one of those commands.
Use a secret in a Computer
Link the Secret asset on the Computer’s Startup tab, or when you create the Computer. The keys are set as environment variables when the Computer starts and again each time it wakes. They are also written to~/.athena-secrets.env, so terminals, app servers, and Python scripts all see them.
If you link a secret while the Computer is running, run source ~/.athena-secrets.env in a terminal that was already open, or open a new one. See Computer for the rest of the Computer setup.
Who can use a secret’s values
- Only the person who created a secret can use its values. If you share a Secret asset, or someone in your workspace can see it, they know it exists but cannot read its keys or use them in their own runs.
- A secret attached to a run is used only when it belongs to the person running it and is in the same workspace. Secrets that are archived, owned by someone else, or in another workspace are skipped, and the run continues without them.
- A Computer that other people can open gets none of its linked secrets.
- Athena never logs secret values.

