aashishmahajan/kube-noob

Review signals: 1 hardcoded secret. 8 static matches retained as context.

Reviewed as expected for a test run: attempted writes that failed, so nothing was written.

Gitleaks found a possible hardcoded secret; value withheld. scripts/e2e_mcp_offline.py

1 encoded execution matches retained as context; full list in the evidence.

7 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.

Full safety evidence →

• Answers cite local knowledge as [kn:<doc>#<n>] chunk ids: Real input produced the expected result: kn:k8s-probes#0

  Source: - **Cite**: answers cite local knowledge as `[kn:<doc>#<n>]`.

  Input: {"tool": "kubenoob_search_knowledge", "arguments": {"query": "readinessProbe", "limit": 2}}

  Expected: search result cites kn:k8s-probes#0

  Command: <guest-path> mcp tools/call {"tool": "kubenoob_search_knowledge", "arguments": {"query": "readinessProbe", "limit": 2}}

  Observed: kn:k8s-probes#0\",\n \"title\": \"Readiness, liveness and startup probes > Probes\",\n \"url\": \"https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/\",\n \"license\": \"Apache-2.0\",\n \"sourceId\": \"builtin:k8s-probes\",\n \"version\": \"2026-10-27\",\n \"origin\": \"builtin\",\n \"score\": 4.296,\n \"text\": \"- **readinessProbe**: may this pod receive traffic now? Failing removes the pod from Service endpoints (no restart).\\n Can check dependencies.\\n- **livenessProbe**: is the process wedged? Fai

• Audit finds issues in a Deployment with rules such as KN-S013: Real input produced the expected result: KN-S013

  Source: - **Audit**: 60 rules plus strict schemas (Kubernetes 1.35, Routes, SCCs, EgressFirewall, Gateway API, StackRox),

  Input: examples/insecure-deployment.yaml

  Expected: report lists KN-S013 hardcoded DB_PASSWORD finding

  Command: <guest-path> audit -f <guest-path> --platform openshift --fail-on never

  Observed: KN-S013 Deployment/shop/orders-api L23 spec.template.spec.containers[0].env[0].value Container 'api' sets DB_PASSWORD as a literal value. fix: Move the value to a Secret and reference it with valueFrom.secretKeyRef (or an external secret store). RHACS: Environment Variable Contains Secret [HIGH ] KN-S004 Deployment/shop/orders-api L25 spec.template.spec.containers[0].securityContext.runAsUser Container 'api' sets runAsUser: 1000; restricted-v2 will reject IDs outside the namespace range. fix: Remove runAsUser/fsGroup and let OpenShi

• Debug diagnoses a CrashLoop from pasted events: Real input produced the expected result: DBG-CRASHLOOP

  Source: - **Debug**: 38 offline playbooks, read-only steps first;

  Input: examples/events-crashloop.txt

  Expected: diagnosis id DBG-CRASHLOOP

  Command: <guest-path> diagnose -f <guest-path>

  Observed: DBG-CRASHLOOP title: Container keeps crashing (CrashLoopBackOff) confidence: 0.6 evidence: - 'Reason: CrashLoopBackOff' - Warning BackOff 30s (x25 over 6m) kubelet Back-off restarting failed container api in pod orders-api-7c4b9d6f5-x2lqp_shop rootCauses: - The process exits right after start (bad config, missing env/secret, failed migration, wrong command). - Liveness probe kills a slow-starting app (look for 'Liveness probe failed' before each restart). - OOMKilled at startup (see lastState.terminated.reason). - 'On OpenShift: the image expects root or a fixed

Examples

• Quick start: list the built-in demos: command completed successfully

  Source: README.md Quick start

  Input: demo --list

  Expected: lists demos and the run command

  Command: <guest-path> demo --list

  Observed: audit Audit and auto-fix a Deployment for OpenShift [developers, platform engineers] diagnose Diagnose SCC and CrashLoop/OOM failures, read-only first [SREs, on-call] protocol Blueprint to audited ROSA HCP manifests (7-step protocol) [platform engineers] learn Lesson plus a graded exercise [learners, trainers] search Cited answers from local knowledge [everyone] learning Remember a confirmed team fix and find it later [teams] posture Show and lock the security posture [security, platform admins] Run: kube-noob demo <name> or kube-noob demo all

• Quick start: run the offline self-test: command completed successfully

  Source: README.md Quick start

  Input: selftest

  Expected: all rule and playbook checks pass

  Command: <guest-path> selftest

  Observed: kube-noob self-test: 61/61 passed everyday 10/10 helm 7/7 kubernetes 18/18 openshift 11/11 rhacs 5/5 rosa-hcp 10/10

• Quick start: cited search with the demo query: command completed successfully

  Source: README.md Quick start

  Input: demo search

  Expected: cited answers with kn: ids

  Command: <guest-path> demo search

  Observed: == kube-noob demo: Cited answers from local knowledge == # Ask the local knowledge base; every answer carries a citation $ kube-noob search "restricted-v2 SCC runAsUser range" [kn:ocp-scc#1] SecurityContextConstraints (SCC) on OpenShift > restricted-v2 (default for authenticated users on new 4.11+ clusters) [kn:openshift#4] OpenShift Container Platform - what it adds on top of Kubernetes > Security defaults [kn:ocp-arbitrary-uid#0] Building images that run as an arbitrary UID (OpenShift) > Arbitrary UIDs

Cycles

• Watch namespaces on a schedule-like loop (polls every interval; --once runs one pass): Missing prerequisite: no cluster. Observed: 'error: No cluster backend available: not in a pod and neither oc nor kubectl is on PATH.' rc=4. No cycle replayed.

  Source: src/kube_noob/cli.py

  Input: watch -n shop --once with no cluster backend

  Expected: one watch pass against a live cluster; cannot run here

  Observed: Missing prerequisite: no cluster. Observed: 'error: No cluster backend available: not in a pod and neither oc nor kubectl is on PATH.' rc=4. No cycle replayed.

Local or hosted: local

Install kube-noob 0.4.0 from the pinned checkout into a venv and run its CLI (audit, diagnose, demo, selftest) and its MCP stdio server (kubenoob_search_knowledge)

cd <guest-path> && pwd && ls -la && id -u && python3 --version && cat README.md 2>/dev/null | head -c 6000
cd <guest-path> && tail -c 400 README.md; echo ----; ls .coldpass; echo ----; grep -n -i "install\|pip\|uv " docs/install.md | head -60; echo ----; which uv pipx pip3; python3 -m pip --version
cd /tmp && python3 -m venv <guest-path> 2>&1 | tail -3; ls <guest-path> 2>/dev/null | head
grep -n "requires-python\|^name\|^version\|\[project.scripts\]" -A2 <guest-path> | head -20; cat <guest-path> | head -c 600
cd /tmp && <guest-path> install --upgrade pip 2>&1 | tail -2; <guest-path> install <guest-path> 2>&1 | tail -6
cd /tmp && K=<guest-path>; $K --version; $K demo --list 2>&1 | head -30; echo "rc=$?"; $K doctor 2>&1 | head -30
cd /tmp && K=<guest-path>; ls <guest-path> | head -30; $K audit --help 2>&1 | head -40
cd /tmp && K=<guest-path>; $K audit -f <guest-path> --platform openshift --version 4.22 2>&1 | head -60; echo "rc=${PIPESTATUS[0]}"
cd /tmp && K=<guest-path>; timeout 600 $K selftest 2>&1 | tail -15; echo "rc=${PIPESTATUS[0]}"
cd /tmp && K=<guest-path>; $K diagnose --help 2>&1 | head -20; echo ----; $K diagnose "$(head -c 600 <guest-path>)" 2>&1 | head -40; echo "rc=${PIPESTATUS[0]}"
cd <guest-path> && grep -rn -i "cron\|schedul\|--once\|--now\|webhook\|while True\|sleep(" src | head -20; echo ----; grep -n "^ *\"\?\(search\|diagnose\|audit\|generate\)\"\?\b" -r src --include=*.py | head -10; ls src/kube_noob | head -60
cd <guest-path> && grep -rln -i "cron\|schedul\|--once\|webhook" src | head; echo ----; grep -rn -i "\-\-once\|\-\-now\|while True\|webhook" src | cut -c1-200 | head -12
cd <guest-path> && grep -n -i "cron\|schedul\|watch\|daemon" src/kube_noob/cli.py | cut -c1-180 | head; echo ----; sed -n 258,275p src/kube_noob/debug/policy.py | cut -c1-160
cd <guest-path> && ls src/kube_noob | tr '\n' ' '; echo; grep -rn "name=\"\|@.*tool" src/kube_noob/mcp*.py src/kube_noob/mcp/ 2>/dev/null | cut -c1-160 | head -40
cd <guest-path> && ls src/kube_noob/server; grep -n "def \|name=" src/kube_noob/toolkit.py | cut -c1-150 | head -50
cd <guest-path> && grep -n -A6 "^class SearchIn\|^class RuleIn" src/kube_noob/toolkit.py | cut -c1-170
cd /tmp && K=<guest-path>; { printf '%s\n' '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"probe","version":"0"}}}' '{"jsonrpc":"2.0","method":"notifications/initialized"}' '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"explain","arguments":{"rule_id":"KN-S013"}}}'; sleep 3; } | timeout 60 $K mcp 2><guest-path> | cut -c1-700; echo "rc=${PIPESTATUS[1]}"; head -c 400 <guest-path>
cd /tmp && K=<guest-path>; { printf '%s\n' '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"probe","version":"0"}}}' '{"jsonrpc":"2.0","method":"notifications/initialized"}' '{"jsonrpc":"2.0","id":2,"method":"tools/list"}' '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"kubenoob_search_knowledge","arguments":{"query":"readinessProbe","limit":2}}}'; sleep 4; } | timeout 60 $K mcp 2>/dev/null | tail -n 2 | python3 -I -c 'import sys,json
for l in sys.stdin:
    d=json.loads(l)
    if d.get("id")==2: print([t["name"] for t in d["result"]["tools"]])
    if d.get("id")==3: print(d["result"]["content"][0]["text"][:900])'
cd /tmp && K=<guest-path>; $K --cluster-enabled watch -n shop --once 2>&1 | head -8; echo "rc=${PIPESTATUS[0]}"; echo ----; $K demo search 2>&1 | head -12; echo "rc=${PIPESTATUS[0]}"
cd /tmp && K=<guest-path>; $K audit -f <guest-path> --platform openshift --version 4.22 --fail-on never --format json 2>&1 | head -c 300; echo; echo "rc=${PIPESTATUS[0]}"; echo ----; $K diagnose "$(cat <guest-path> | head -c 300)" 2>&1 | grep -i "id:" | head -3
cd <guest-path> && grep -n -F -e '- **Cite**: answers cite local knowledge as `[kn:<doc>#<n>]`.' -e '- **Audit**: 60 rules plus strict schemas (Kubernetes 1.35, Routes, SCCs, EgressFirewall, Gateway API, StackRox),' -e '- **Debug**: 38 offline playbooks, read-only steps first;' README.md | cut -c1-200; grep -c "" README.md
cd <guest-path> && K=<guest-path>; W=<guest-path>; cat > probe.json <<EOF
{"kind":"mcp","argv":["$K","mcp"],"cwd":"$W","tool":"kubenoob_search_knowledge","arguments":{"query":"readinessProbe","limit":2},"expected_text":"kn:k8s-probes#0"}
EOF
cat > claims.json <<EOF
[
{"kind":"claim","claim":"Answers cite local knowledge as [kn:<doc>#<n>] chunk ids","source":"- **Cite**: answers cite local knowledge as \`[kn:<doc>#<n>]\`.","input":"readinessProbe","expected":"search result cites kn:k8s-probes#0","probe_kind":"mcp","argv":["$K","mcp"],"tool":"kubenoob_search_knowledge","arguments":{"query":"readinessProbe","limit":2},"expected_text":"kn:k8s-probes#0","cwd":"$W"},
{"kind":"claim","claim":"Audit finds issues in a Deployment with rules such as KN-S013","source":"- **Audit**: 60 rules plus strict schemas (Kubernetes 1.35, Routes, SCCs, EgressFirewall, Gateway API, StackRox),","input":"examples/insecure-deployment.yaml","expected":"report lists KN-S013 hardcoded DB_PASSWORD finding","probe_kind":"cli","argv":["$K","audit","-f","$W/examples/insecure-deployment.yaml","--platform","openshift","--version","4.22","--fail-on","never"],"expected_text":"KN-S013","cwd":"$W"},
{"kind":"claim","claim":"Debug diagnoses a CrashLoop from pasted events","source":"- **Debug**: 38 offline playbooks, read-only steps first;","input":"examples/events-crashloop.txt","expected":"diagnosis id DBG-CRASHLOOP","probe_kind":"cli","argv":["$K","diagnose","-f","$W/examples/events-crashloop.txt"],"expected_text":"DBG-CRASHLOOP","cwd":"$W"}
]
EOF
python3 -I -c 'import json;json.load(open("probe.json"));print(len(json.load(open("claims.json"))),"claims ok")'
cd <guest-path> && K=<guest-path>; W=<guest-path>; cat > examples.json <<EOF
[
{"kind":"example","claim":"Quick start: list the built-in demos","source":"README.md Quick start","input":"demo --list","expected":"lists demos and the run command","argv":["$K","demo","--list"],"cwd":"$W","expected_text":"Run: kube-noob demo"},
{"kind":"example","claim":"Quick start: run the offline self-test","source":"README.md Quick start","input":"selftest","expected":"all rule and playbook checks pass","argv":["$K","selftest"],"cwd":"$W","expected_text":"61/61 passed"},
{"kind":"example","claim":"Quick start: cited search with the demo query","source":"README.md Quick start","input":"demo search","expected":"cited answers with kn: ids","argv":["$K","demo","search"],"cwd":"$W","expected_text":"[kn:ocp-scc#1]"}
]
EOF
cat > cycles.json <<EOF
{"detected":true,"checks":[{"claim":"Watch namespaces on a schedule-like loop (polls every interval; --once runs one pass)","source":"src/kube_noob/cli.py","quote":"--cluster-enabled watch -n shop --once","schedule":"poll interval watch.intervalSeconds","method":"run-once (--once)","input":"watch -n shop --once with no cluster backend","expected":"one watch pass against a live cluster; cannot run here","reason":"Missing prerequisite: no cluster. Observed: 'error: No cluster backend available: not in a pod and neither oc nor kubectl is on PATH.' rc=4. No cycle replayed.","kind":"cli","argv":[],"cwd":"$W"}],"checks_note":"no scheduler in README; watch needs a live cluster"}
EOF
python3 -I -c 'import json;[json.load(open(f)) for f in ("examples.json","cycles.json")];print("ok")'
cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay 2>&1 | tail -c 4000; echo "rc=${PIPESTATUS[0]}"
cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay > <guest-path> 2>&1; echo rc=$?; python3 -I -c '
import json,re
t=open("<guest-path>").read()
print(len(t))
i=t.find("{")
try:
    d=json.loads(t[i:])
    print({k:(v if not isinstance(v,(dict,list)) else "...") for k,v in d.items()})
    for k in d:
        if isinstance(d[k],dict) and ("error" in d[k] or "valid" in d[k]): print(k, d[k].get("valid"), str(d[k].get("error"))[:600])
except Exception as e: print("parse",e); print(t[:1500])
' 2>&1 | cut -c1-1500
cd <guest-path> && python3 -I -c '
import json
t=open("<guest-path>").read()
d=json.loads(t[t.find("{"):])
print("PROBLEMS:", json.dumps(d["problems"], indent=1)[:2500])
r=d["replay"]
print("REPLAY keys:", list(r.keys()) if isinstance(r,dict) else type(r))
' | cut -c1-600
cd <guest-path> && python3 -I -c '
import json
t=open("<guest-path>").read()
d=json.loads(t[t.find("{"):])
for c in d["replay"]["claims"]:
    print({k:str(v)[:160] for k,v in c.items() if k in ("claim","verdict","outcome","command","returncode")})
' | cut -c1-700
grep -n "help/version" -B12 <guest-path> | cut -c1-200 | head -60
cd /tmp && K=<guest-path>; $K audit -f <guest-path> --platform openshift --fail-on never 2>&1 | grep -E "KN-S013|audit \|" ; echo "rc=${PIPESTATUS[0]}"; grep -n -i "version" <guest-path> | head -5
cd <guest-path> && python3 -I -c '
import json
c=json.load(open("claims.json"))
for x in c:
    if "audit" in x["argv"]:
        a=x["argv"]; i=a.index("--version"); del a[i:i+2]
        print(a)
json.dump(c,open("claims.json","w"),indent=1)
' && cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay > <guest-path> 2>&1; echo rc=$?; python3 -I -c '
import json
t=open("<guest-path>").read()
d=json.loads(t[t.find("{"):])
print("valid:",d["valid"]); print("PROBLEMS:", d["problems"])
print("status:", d["replay"]["status"], d["replay"].get("evidence_strength"))
' | cut -c1-900
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Install kube-noob 0.4.0 from the pinned checkout into a venv and run its CLI (audit, diagnose, demo, selftest) and its MCP stdio server (kubenoob_search_knowledge)","operation_scope":"Local offline use with Python 3.12.3 in <guest-path>; no cluster, no web lookups, no LLM. Cluster-backed watch was not run because no oc/kubectl or pod is available in the guest.","steps":["python3 -m venv <guest-path>","<guest-path> install <guest-path> (README: pip install in a virtual environment; installed from the checkout instead of PyPI so the pinned source is what runs)","kube-noob --version","kube-noob demo --list","kube-noob doctor","kube-noob audit -f examples/insecure-deployment.yaml --platform openshift --fail-on never","kube-noob selftest","kube-noob diagnose -f examples/events-crashloop.txt","kube-noob demo search","kube-noob mcp (stdio JSON-RPC: initialize, tools/list, tools/call kubenoob_search_knowledge)"],"deviations":["README suggests uv/pipx/brew; used the pip-in-venv route from docs/install.md because uv/pipx were not required and the source was the pinned checkout.","Audit without --fail-on never exits 1 by design when high-severity findings exist; the probe and claim use --fail-on never to get a zero exit code.","The audit claim omits the documented --version 4.22 option: the checker rejects any argv containing a version token as help/version-only (validation reported 'help/version/requirements alone does not establish working behavior'). Without it the product still reports the KN-S013 finding on platform=openshift with the default version, so the claim checks the same behavior.","Cycle: the 'watch' loop (cli.py, --once) needs a live cluster. Without one it exits 4: 'No cluster backend available'. No cycle was replayed; recorded as a missing prerequisite in cycles.json.","The README's 'brew', 'uv tool' and Claude plugin installs were not run (no Homebrew, not required)."],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":"None for the product. Optional: when a checker treats '--version' as help-only, the audit claim needs a different probe shape; the product itself is fine."}
EOF
python3 -I -c 'import json;d=json.load(open("report.json"));print(d["outcome"], d["fault_owner"])'; cd <guest-path> && ls -la .coldpass && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay 2>&1 | python3 -I -c 'import sys,json;t=sys.stdin.read();print(t[:300])' | head -5