Files
marqdo-cli/docs/evidence/2026-11-20-strict-gc-8af7262.md
cflmy a81509dadd docs(evidence): bind the G-C and RC reports to the final code SHA
The convergence evidence was first recorded against f075c96. The completed
tree is 8af7262 (docs/workflow on top of the same code), and the G-C lane and
workspace suite were re-run from a clean `git archive 8af7262`. Rename the
evidence and RC reports to the final SHA and update their archive/checksum
hashes.

Source under test: 8af72620f1
G-C: 7/7 + positive CLI + forged-marker rejection PASS (ordinary user, VM)
Workspace: 47 test binaries, 0 failed
RC: marqdo-0.1.0-x86_64-unknown-linux-gnu.tar.gz,
    sha256 5a42d83737c9715986f36b105feb7419a677d98f913f00e84ac19f1d4847733d
CI_URL: none from this host; remote gates remain the blockers.
2026-11-20 15:58:59 +08:00

4.7 KiB

G-C strict-isolation evidence — source archive 8af7262 (2026-11-20)

Source SHA: 8af72620f12e2494a59830a23abb393768b610e9 Source archive: git archive --format=tar.gz --prefix=marqdo-8af7262/ 8af7262 SHA-256 ad8fca0b5e114b0c74d6265e5e374373a50bccd6a58aa12f821855cf2c9ed7ea.

Status: G-C local VM lane PASS, run as ordinary user dev (uid 1000) in a disposable Ubuntu 26.04 QEMU/TCG VM. No self-hosted GitHub Actions security URL exists for this SHA, so this is not yet a CI-signed G-C.

Full logs: 2026-11-20-strict-gc-8af7262.log, 2026-11-20-workspace-tests-8af7262.log.

Environment

id:      uid=1000(dev) gid=1000(dev) groups=1000(dev),27(sudo)
uname:   Linux marqdo-security 7.0.0-43-generic #43-Ubuntu SMP ... x86_64
virt:    systemd-detect-virt --container -> none (QEMU/TCG guest)
bwrap:   bubblewrap 0.11.1 (bwrap-userns-restrict)
cc:      cc (Ubuntu 15.2.0-16ubuntu1) 15.2.0
node:    v22.22.1
python:  Python 3.14.4
rustc:   rustc 1.100.0 (0ef1bc386 2026-11-09)
uid_map: 0 0 4294967295
cgroup2: rw,nosuid,nodev,noexec,relatime,nsdelegate,...

The lane was extracted from git archive, not run from the development container. CARGO_TARGET_DIR pointed at the cached VM target directory; the lane was fixed to honour that variable rather than hard-coding target/debug.

Lane command

cd ~/marqdo-8af7262
export PATH=$HOME/.cargo/bin:$PATH
export CARGO_TARGET_DIR=$HOME/marqdo-src/target
MARQDO_REQUIRE_STRICT_SECURITY=1 ./tests/security/run-lane.sh

Result

== adversarial helper matrix ==
running 7 tests
test strict_child_cannot_escape_via_proc_or_privileges ... ok
test strict_child_cannot_reach_the_network ... ok
test strict_child_cannot_read_host_files_but_reads_the_project ... ok
test strict_child_cannot_write_outside_or_into_the_project ... ok
test strict_child_identity_is_unprivileged_and_namespaced ... ok
test strict_supervisor_enforces_fd_and_process_limits ... ok
test strict_supervisor_enforces_memory_cpu_and_output_limits ... ok
test result: ok. 7 passed; 0 failed; 0 ignored

strict mechanism: bubblewrap 0.11.1 (user/mount/PID/net/UTS/IPC namespaces + no-new-privs + rlimits)
== positive CLI contract ==
== external-execution surfaces are refused ==
== forged active marker is rejected ==
G-C security lane: PASS

The same archive ran cargo test --workspace --locked: 47 test binaries, 0 failed, including the seven security cases in parallel (the parallel scratch-path collision is fixed in this SHA).

Case-level observations

Attack/surface Observed inside sandbox
identity uid=1000, pid=1, NoNewPrivs: 1, CapEff: 0
/etc/shadow, /root, $HOME DENIED
/project/allowed.txt OK
write outside project / into /project DENIED; no host file created
write /tmp allowed in private tmpfs, host unaffected
/proc/1/root/etc/shadow DENIED
inherited fds above stdio none
setuid(0), mount, nested unshare DENIED
IPv4 connect / raw socket DENIED
fd loop EMFILE <= 64
fork loop bounded below 200
768 MiB allocation denied/killed by 512 MiB RLIMIT_AS
infinite CPU loop stopped by wall-clock/CPU budget
4 MiB stdout stopped at the 64 KiB output limit

CLI behavior in the VM

run examples/hello.mq.md --mode strict                -> 你好,Marqdo (exit 0)
call examples/factorial.mq.md 数学.阶乘 --arg n=5 --mode strict -> 120
run examples/multi-module/index.mq.md --mode strict   -> exit 0
run --mode strict --json                              -> success: true
run --mode strict --allow-foreign                     -> exit 2
run --mode strict --allow-read/--allow-write          -> exit 2
run --mode strict --trust-native-plugin --plugin ...  -> exit 2
run --mode strict --kernel ...                        -> exit 2
run --mode strict --trace-file ...                    -> exit 2
MARQDO_STRICT_ACTIVE=1 run --mode strict              -> exit 1 (not PID 1)

Residual risks and non-claims

  • Strict L2 (Worker/Foreign/native ABI inheriting the sandbox) is not implemented; those flags are rejected. Native ABI is never allowed in strict.
  • rlimits are used instead of cgroup-v2 accounting; no cgroup enforcement is claimed.
  • Filesystem TOCTOU is reduced by read-only mounts but not formally proven against a racing host user.
  • Network is denied by the network namespace, not a seccomp syscall allowlist.
  • A remote self-hosted security-lane CI run is still required before a release may call G-C CI-verified. .github/workflows/security-linux.yml defines that lane; MARQDO_REQUIRE_STRICT_SECURITY=1 makes a missing mechanism fail instead of skip.