Haiku 5.5
Native agent · Checked operation passedChecked operation go run ./examples/minimal
Agent-reported scope: Run the README quickstart: SANDBOX_PROVIDER=wasm go run ./examples/minimal, which executes one JavaScript snippet in the in-process wasm sandbox and prints the department totals.
Run details
- run
- plimsollmark--plimsoll-claude-haiku-5-5-1793145239219658611
- date
- Oct 27, 2026
- time to working
- 153.0s
- product hosting / execution
- Local command in the guest
- recorded operation scope
- product operation
- model
- claude-haiku-5-5
- agent / version
- claude-code / 2.1.310 (Claude Code)
- platform
- linux
- reported cost
- $0.011477
- Recorded source
- https://github.com/plimsollmark/plimsoll
- Backend
- hyperv
- Tool transport
- mcp-assigned-vm
- Requested source pin (not an observed revision)
- {"commit": "e543f3c78de95cdfbe659774cf72efdb9a176829", "url": "https://github.com/plimsollmark/plimsoll"}
- Test version
- readme-d5b140d3c0f8b95c2dd6fe6fc381a303cacb92679582c22317d8e78b572f0cf2
- revision
- e543f3c78de95cdfbe659774cf72efdb9a176829
Safety: Review signals: 4 encoded blob, 1 hardcoded secret, 195 known vulnerability. 3 static matches retained as context.
Review signals: 4 encoded blob, 1 hardcoded secret, 195 known vulnerability. 3 static matches retained as context.
Source contains encoded blob; review its purpose. docs/trainers/mcp-tour.js
Gitleaks found a possible hardcoded secret; value withheld. docs/callers.md
OSV reports a known vulnerability in this dependency version. 27 distinct locations. 189 findings.
3 hardcoded secret matches retained as context; full list in the evidence.
OSV scanned supported lockfiles; unpinned dependencies and unsupported formats are outside this check.
A short run cannot establish that code is safe.
Connections show IP addresses; proxies can hide the final hostname.
Tracing covers dispatched commands and their children, not the whole OS; guest root can interfere with tracing.
Only 200 of 316 findings are stored; the summary counts all of them.
Answers recorded for this run
- The minimal example runs without a daemon, Docker, credentials or network: Real input produced the expected result: "Sales":20300000
Source: The example's execution needs no daemon, Docker, credentials or outbound network.
Input: go run ./examples/minimal with SANDBOX_PROVIDER=wasm
Expected: JSON totals per department printed by the sandboxed snippet
Command: go run ./examples/minimal
Observed: "Sales":20300000,"Operations":9800000}
- The example explicitly selects a provider: Real input produced the expected result: provider wasm
Source: The example explicitly selects a provider.
Input: go run ./examples/minimal with SANDBOX_PROVIDER=wasm
Expected: provider line reports wasm
Command: go run ./examples/minimal
Observed: provider wasm isolation process exit code 0 timed out false truncated stdout=false stderr=false duration 359ms stdout {"Engineering":59000000,"Sales":20300000,"Operations":9800000}
Examples
- Quickstart: run the minimal sandbox example with the wasm provider: command completed successfully
Source: README.md#run-something-locally
Input: SANDBOX_PROVIDER=wasm go run ./examples/minimal
Expected: provider wasm, isolation process, exit code 0, stdout with department totals
Command: go run ./examples/minimal
Observed: provider wasm isolation process exit code 0 timed out false truncated stdout=false stderr=false duration 346ms stdout {"Engineering":59000000,"Sales":20300000,"Operations":9800000}
Local or hosted: both
What the agent did
Run the README quickstart: SANDBOX_PROVIDER=wasm go run ./examples/minimal, which executes one JavaScript snippet in the in-process wasm sandbox and prints the department totals.
cd <guest-path> && pwd && ls -la && git status 2>&1 | head -3; id -u; python3 --version; cat /etc/os-release | head -2; echo "CDP=$CDP_URL"; ls /inputs 2>&1
cd <guest-path> && ls README* && wc -c README* && head -c 7000 README.md
cd <guest-path> && tail -c +7000 README.md | head -c 6000
cd <guest-path> && cat go.mod | head -8; echo ---; cat toolchain.versions; echo ---; which go; go version 2>&1; ls /usr/local/go 2>&1 | head -2
cd <guest-path> && go version && time (SANDBOX_PROVIDER=wasm timeout 900 go run ./examples/minimal > /tmp/minimal.out 2>&1; echo "exit=$?") ; tail -c 3000 /tmp/minimal.out
cd <guest-path> && tail -c +13000 README.md | head -c 6400
cd <guest-path> && grep -n -i -E "cron|schedul|--once|--now|webhook|ticker|loop" README.md | head; echo ---; ls examples | head -40; echo ---; ls docs | head -40; echo; ls cmd
cd <guest-path> && grep -rn -E "time\.NewTicker|time\.Tick\(|cron" cmd/plimsolld examples/minimal 2>/dev/null | head -5; echo "rc=$?"; ls -la .coldpass; cat examples/minimal/*.go | head -60
cd <guest-path> && sed -n 645,675p cmd/plimsolld/main.go
cd <guest-path> && grep -rln -E "func \(.*\) (Preflight|Reconcile[A-Za-z]*)\(" sandbox | head -20; echo ---; ls sandbox | head -40
cd <guest-path> && grep -n -E "func \(.*\) (Preflight|Reconcile[A-Za-z]*)\(" sandbox/wasm.go; grep -n -E "OrphanReconciler|Preflighter" sandbox/*.go | grep -v _test | head -8; grep -n -B2 -A12 "func (.*) Preflight(" sandbox/wasm.go | head -30
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["go","run","./examples/minimal"],"cwd":"<guest-path>","env":{"SANDBOX_PROVIDER":"wasm"},"expected_text":"\"Sales\":20300000"}
EOF
cat > claims.json <<'EOF'
[{"kind":"claim","claim":"The minimal example runs without a daemon, Docker, credentials or network","source":"The example's execution needs no daemon, Docker, credentials or outbound network.","input":"go run ./examples/minimal with SANDBOX_PROVIDER=wasm","expected":"JSON totals per department printed by the sandboxed snippet","probe_kind":"cli","argv":["go","run","./examples/minimal"],"env":{"SANDBOX_PROVIDER":"wasm"},"expected_text":"\"Sales\":20300000","cwd":"<guest-path>"},
{"kind":"claim","claim":"The example explicitly selects a provider","source":"The example explicitly selects a provider.","input":"go run ./examples/minimal with SANDBOX_PROVIDER=wasm","expected":"provider line reports wasm","probe_kind":"cli","argv":["go","run","./examples/minimal"],"env":{"SANDBOX_PROVIDER":"wasm"},"expected_text":"provider wasm","cwd":"<guest-path>"}]
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Quickstart: run the minimal sandbox example with the wasm provider","source":"README.md#run-something-locally","input":"SANDBOX_PROVIDER=wasm go run ./examples/minimal","expected":"provider wasm, isolation process, exit code 0, stdout with department totals","argv":["go","run","./examples/minimal"],"cwd":"<guest-path>","env":{"SANDBOX_PROVIDER":"wasm"},"expected_text":"\"Sales\":20300000"}]
EOF
cat > cycles.json <<'EOF'
{"detected":false,"reason":"README has no cron, schedule, --once, --now or webhook behaviour. The only loop found is cmd/plimsolld/main.go (time.NewTicker(time.Minute), ~line 658), a daemon-internal isolation re-check that logs only on a failed preflight; the wasm Preflight returns nil so a successful cycle has no observable output. The README quickstart does not start plimsolld, so no cycle was triggered.","checks":[]}
EOF
for f in *.json; do python3 -I -m json.tool "$f" >/dev/null && echo "ok $f"; done
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Run the README quickstart: SANDBOX_PROVIDER=wasm go run ./examples/minimal, which executes one JavaScript snippet in the in-process wasm sandbox and prints the department totals.","operation_scope":"Local only. Needs Go (installed 1.26.10 via the module toolchain line; /usr/bin/go on Ubuntu 24.04). No daemon, Docker, credentials or outbound network needed for the run itself. Go module downloads needed network on first run. Daemon plimsolld and gVisor/Docker/hosted providers were not exercised.","steps":["SANDBOX_PROVIDER=wasm go run ./examples/minimal"],"deviations":["Used the README's installed Go (/usr/bin/go, version go1.26.10 downloaded by the toolchain line) instead of a separate install step; no install was needed.","Did not start plimsolld or run the Docker/gVisor/hosted providers; the README quickstart does not require them."],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":"None for the quickstart. The daemon's one-minute isolation re-check was not triggered and is not covered by this run."}
EOF
python3 -I -m json.tool report.json >/dev/null && echo ok; cd <guest-path> && <env-var>=<guest-path> timeout 900 python3 -I <guest-path> --validate --replay 2>&1 | tail -c 4000run JSON · checker evidence · agent report + check · transcript excerpt · error excerpt unavailable