Merge pull request 'Document the agent-host access model' (#17) from docs/host-access-model into main

Reviewed-on: #17
This commit was merged in pull request #17.
This commit is contained in:
inference-admin committed 2026-11-21 16:14:01 +00:00
commit 8d002bf5bc
2 files changed
+32 -5

No files matched your search

+21 -2
View File
@@ -19,6 +19,24 @@ This page is for members who read its output or give it instructions.
hyphen) no longer holds an access token; some historical references and old
room roles still carry the name.
## Who can do what on the host
Each agent runs as its own Unix account, and the accounts are ranked:
- **The bots** (`infrastructure-bot`, `socialbot`, `designbot` and any future
working-group bot) cannot use `sudo` at all.
- **Inferencebot, the steward,** can act inside those bot accounts — it is the
fleet administrator — but it has **no root** and no passwordless `sudo` beyond
the docs and backup service commands. It cannot install packages, create users,
edit `/etc`, or open a root shell.
- **Root belongs to a person.** A separate operator account, reached with an SSH
key and never a password, holds full `sudo`. Root SSH login stays closed: the
operator signs in as their own account and escalates deliberately.
That split is deliberate. An agent compromise reaches the bot accounts and the
co-op plumbing, not the host. Host-level actions stay attributable to a human,
and automation cannot quietly widen its own privileges.
## What to ask for
- Operational work that a runbook already describes, or that ends in a PR, an
@@ -57,8 +75,9 @@ operator can stop a deploy; anything irreversible waits for confirmation.
the chat: confirmed before they deploy.
- Changing the model catalog, prices or budgets.
- Minting or rotating credentials.
- Host root actions outside the docs/backup service allowlist: package installs,
user management, and arbitrary root commands.
- Host root actions — package installs, user management, sudoers changes, and
arbitrary root commands. These run only through the operator's key-only
account; the steward cannot perform them.
- Anything a runbook marks "confirm".
This matches the standing policy the agents operate under today. A stricter
+11 -3
View File
@@ -38,13 +38,19 @@ and deliberately kept apart from the member-facing applications. The steward bot
— the automation behind the digests and the newsletter drafts — and the co-op's
other agents live there. Keeping them on their own host means an agent problem
stays an agent problem: it cannot reach the chat, the gateway or member data.
Access to that host is SSH-key only — no agent or bot account has a password.
The steward can act inside the bot accounts it maintains but cannot take root;
root is a separate human operator account, so host-level changes stay with a
person.
### Backups
Cloudron writes **encrypted** backups to Infomaniak's object storage in
Switzerland every night and keeps the last **seven days**, so a failed upgrade
or a lost server can be restored. The automation host keeps its own encrypted
backup as well, with credentials kept apart from the agent's environment.
backup as well, with credentials kept apart from the agent's environment. That
backup runs as root from a root-owned script, so the agent can trigger it but
cannot alter the backup code or read the backup keys.
### Domain and DNS
@@ -215,8 +221,10 @@ private, loopback, and internal addresses. The
independent and independently rotatable, so a leak has bounded scope.
- **No secrets in version control.** Credentials live in the hosting platform's
environment, never in the repos.
- **Least-privilege automation.** The steward agent runs as its own Unix user;
host-level actions that need root are operator-controlled.
- **Least-privilege automation.** Every agent runs as its own Unix user; bots have
no `sudo`, the steward can act only within the bot accounts it maintains, and
root belongs to a separate key-only operator account. Host-level changes are
therefore deliberate human actions, not automation.
- **Accepted trade-offs** (documented, not hidden): member keys are stored in
the portal database so the co-op can serve them, and the API gateway must stay
publicly reachable. Its admin surfaces are disabled; only the public API is