fix(docker): prevent privilege regain after bootstrap

Co-authored-by: yu-xin-c <175149126+yu-xin-c@users.noreply.github.com>

.github/workflows/ci.yml

@@ -189,10 +189,47 @@ jobs:
       - name: Build image with default channel dependencies
         run: docker build -t nanobot:test .
 
-      - name: Verify Docker Compose startup
+      - name: Verify Docker Compose startup and privilege boundary
         env:
           HOME: ${{ runner.temp }}
-        run: docker compose run --rm --no-deps --build -T nanobot-cli status
+        run: |
+          docker compose run --rm --no-deps --build -T nanobot-cli status
+          docker compose run --rm --no-deps -T --entrypoint sh nanobot-cli -s <<'OUTER'
+          set -eu
+          field() {
+            awk -v key="$1:" '$1 == key { print $2 }' /proc/self/status
+          }
+          test "$(id -u)" = "0"
+          test "$(field NoNewPrivs)" = "1"
+          setpriv --reuid=nanobot --regid=nanobot --init-groups sh -s <<'INNER'
+          set -eu
+          field() {
+            awk -v key="$1:" '$1 == key { print $2 }' /proc/self/status
+          }
+          test "$(id -u)" = "1000"
+          test "$(field NoNewPrivs)" = "1"
+          for capability_set in CapInh CapPrm CapEff CapAmb; do
+            test "$(field "$capability_set")" = "0000000000000000"
+          done
+          INNER
+          OUTER
+          docker compose -f docker-compose.yml -f docker-compose.bwrap.yml \
+            run --rm --no-deps -T --entrypoint sh nanobot-cli -s <<'BWRAP'
+          set -eu
+          mkdir -p /home/nanobot/.nanobot/workspace
+          chown -R nanobot:nanobot /home/nanobot/.nanobot
+          setpriv --reuid=nanobot --regid=nanobot --init-groups \
+            bwrap --new-session --die-with-parent \
+            --ro-bind /usr /usr \
+            --ro-bind-try /bin /bin \
+            --ro-bind-try /lib /lib \
+            --ro-bind-try /lib64 /lib64 \
+            --proc /proc --dev /dev --tmpfs /tmp \
+            --bind /home/nanobot/.nanobot/workspace /home/nanobot/.nanobot/workspace \
+            --chdir /home/nanobot/.nanobot/workspace \
+            -- sh -c 'test "$(id -u)" = 1000; touch sandbox-ok'
+          test -f /home/nanobot/.nanobot/workspace/sandbox-ok
+          BWRAP
 
       - name: Verify default WhatsApp dependencies
         run: docker run --rm --entrypoint python nanobot:test -c "import neonize, segno"

docker-compose.yml

@@ -13,6 +13,10 @@ x-common-config: &common-config
     - CHOWN
     - SETGID
     - SETUID
+  # Prevent the non-root process from regaining capabilities through setuid
+  # binaries or file capabilities left inside the container image.
+  security_opt:
+    - no-new-privileges:true
 
 services:
   nanobot-gateway:

docs/deployment.md

@@ -160,8 +160,11 @@ docker compose logs -f nanobot-gateway                   # view logs
 docker compose down                                      # stop
 ```
 
-The default Compose file drops all Linux capabilities and keeps Docker's default
-AppArmor/seccomp profiles enabled. If you explicitly set
+The default Compose file drops all Linux capabilities except `CHOWN`, `SETUID`, and
+`SETGID`, which the root entrypoint needs to fix bind-mount ownership and become UID
+1000. It also enables `no-new-privileges`, so the non-root process cannot regain those
+bootstrap capabilities through setuid binaries or file capabilities. Docker's default
+AppArmor/seccomp profiles remain enabled. If you explicitly set
 `"tools.exec.sandbox": "bwrap"` in `~/.nanobot/config.json`, add the bwrap
 override file when starting containers:
 
@@ -170,9 +173,9 @@ docker compose -f docker-compose.yml -f docker-compose.bwrap.yml up -d nanobot-g
 docker compose -f docker-compose.yml -f docker-compose.bwrap.yml run --rm nanobot-cli agent -m "Hello!"
 ```
 
-The override grants `CAP_SYS_ADMIN` and disables AppArmor/seccomp confinement for
-the container so bubblewrap can create its nested namespaces. Use it only when the
-bwrap sandbox is enabled.
+The override adds `CAP_SYS_ADMIN` and disables AppArmor/seccomp confinement for the
+container so bubblewrap can create its nested namespaces. It preserves
+`no-new-privileges`. Use it only when the bwrap sandbox is enabled.
 
 ### Docker
 
@@ -197,6 +200,8 @@ vim ~/.nanobot/config.json
 # health endpoint on 18790.
 docker run \
   --cap-drop ALL \
+  --cap-add CHOWN --cap-add SETGID --cap-add SETUID \
+  --security-opt no-new-privileges:true \
   -v ~/.nanobot:/home/nanobot/.nanobot \
   -p 18790:18790 -p 8765:8765 \
   nanobot gateway
@@ -205,7 +210,9 @@ docker run \
 # bubblewrap needs for nested namespaces. Without them, `bwrap` may exit with
 # `clone3: Operation not permitted`.
 docker run \
-  --cap-drop ALL --cap-add SYS_ADMIN \
+  --cap-drop ALL \
+  --cap-add CHOWN --cap-add SETGID --cap-add SETUID --cap-add SYS_ADMIN \
+  --security-opt no-new-privileges:true \
   --security-opt apparmor=unconfined \
   --security-opt seccomp=unconfined \
   -v ~/.nanobot:/home/nanobot/.nanobot \