Home-Network/home-network-k3s-impl
2
0 Fork 0
Code Issues 2 Pull Requests 5 Actions Packages Projects Releases Wiki Activity

feat(vuln): automated CVE visibility for production (R-VULN-*, ADR-0027) #165

Merged
jakman85 merged 18 commits from feat/vuln-visibility into main 2026-11-09 22:41:50 +00:00
Conversation 0 Commits 18 Files Changed 99
+9295 -9
gitea-admin commented 2026-11-09 18:41:24 +00:00
Owner
Copy link
Copy source

Implements docs/17-vulnerability-visibility.md (ADR-0027) in full, R-VULN-01..09: a scheduled in-cluster Trivy scanner, an SSO-gated CVE page, aggregate-only metrics and alerts, and a Renovate-shaped triage bot.

Depends on spec PR Home-Network/home-network-k3s#4, merge that first; SPEC_PIN not bumped. The merge SHA does not exist yet, and open PR #160 owns the pin backlog.

Requirements and spec sections served

R-ID Spec section Where
R-VULN-01 Inventory platform/vuln-scanner/src/inventory.py
R-VULN-02 Scanning src/main.py, src/trivy.py
R-VULN-03 Coverage src/state.py, src/report.py
R-VULN-04 The CVE page src/page.py, sidecar, HTTPRoute (login blocked, see operator action)
R-VULN-05 Trigger, no human step scan loop, .gitea/workflows/cve-triage.yaml
R-VULN-06 Alerting platform/vuln-scanner/alerts.yaml, src/metrics.py, promtool tests
R-VULN-07 The triage bot .gitea/workflows/scripts/cve_triage.py
R-VULN-08 Risk acceptance platform/vuln-scanner/acceptances.yaml, src/acceptances.py, L0 check
R-VULN-09 Security of the scanner rbac.yaml, networkpolicy.yaml, digest pins, renovate.json5

Also surfaces R-OBS-02/03, R-C2-02, R-UPD-01/02, R-SEC-02, R-IDP-06.

Own Flux tier (clusters/base/vuln-scanner.yaml): wait: false, dependsOn: platform-config, nothing depends on it, so this tier cannot stall the platform.

Gate

Local toolbox gate on e1adc39: rc 3, UNDETERMINED, not green.

[skip] stage spec-immutability: SKIPPED (not run)
[skip] stage gitea-settings: SKIPPED (not run)
[skip] stage conformance: SKIPPED (not run)
[ ok ] approval-policy: green
[ ok ] decisions-log: green
[ ok ] lint: green
[ ok ] l0: green
[ ok ] render-golden: green
[ ok ] l1: green
[fail]   Stages that did not run: spec-immutability gitea-settings

conformance (L2/L3) skipped: no cluster. approval-policy is N/A offline and proves nothing locally. L2/L3 tests for this feature are scaffolds that do not run in CI.

Renovate validator (renovate:44.167.0, the tag the workflow runs): validated successfully.

Operator action required

  1. Review the sensitive paths. This PR touches .gitea/**, conformance/**, clusters/** and RBAC, so approval-policy is expected to be red in CI until an administrator reviews. That is the correct stopping point. No agent will post /approve-override.
  2. Add the Keycloak client. R-VULN-04 is NOT delivered until this is done. Login fails closed (invalid_client) until then; the scanner, metrics and alerts are unaffected. An agent cannot do it: the client secret must be a new key in the existing platform/keycloak/keycloak-realm-clients.sops.yaml (needs the age private key), and the live realm needs a one-time kcadm client creation. Exact ordered steps are in platform/vuln-scanner/README.md, "Operator action required". Those steps are UNVERIFIED.

Expected after deploy

  • VulnCriticalFixAvailable will fire on day one. That is the backlog becoming visible, as ADR-0027 says.
  • VulnTriageBotSilent fires until the bot's first successful run at 07:15 UTC or a workflow_dispatch run.
  • VulnScanStale fires during the first full scan cycle, because the last-complete metric is absent until one completes.

Spec escalations

  1. credential-required is defined as "the registry refused anonymous access", but the egress contract excludes LAN ranges, so the scanner cannot reach the private registry to be refused. Implemented by naming private registries in VULN_PRIVATE_REGISTRIES, short-circuiting before Trivy.
  2. The egress contract lists only the Kubernetes API, DNS and public TCP/443, but the SSO sidecar must reach Keycloak. Egress additionally allows the Gateway on 443 and Keycloak.

Deviations from the recommended shape

  • Page is server-rendered with no JavaScript rather than inline JS: no script-src in the CSP, one escaping path, and L1 tests exercise the real rendered page.
  • Three pod ports (page, internal, metrics) so Prometheus cannot read report.json.
  • Python 3.14.9-alpine, the version the tests ran on; Trivy 0.76.1, as 0.76.2 was not old enough.

UNVERIFIED until it runs in production

Nothing has been deployed. No runtime behaviour has been observed.

  • A real Trivy run; every scanner test uses a recorded fixture and a fake runner.
  • The real Kubernetes API, browser rendering, OIDC login and the platform-admins restriction.
  • In-cluster DNS from a DinD job container, and reachability of the pinned ClusterIP fallback 10.43.210.10. The workflow has never run; Gitea behaviour is tested only against a fake.
  • Renovate's ## Pending Approval row format, because issue #91 has no live approve row today.
  • minimumReleaseAge behaviour on a real update.
  • Cgroup v2 memory behaviour under the scanner's limits (inferred).

🤖 Generated with Claude Code

Implements `docs/17-vulnerability-visibility.md` (ADR-0027) in full, R-VULN-01..09: a scheduled in-cluster Trivy scanner, an SSO-gated CVE page, aggregate-only metrics and alerts, and a Renovate-shaped triage bot. **Depends on spec PR Home-Network/home-network-k3s#4, merge that first; SPEC_PIN not bumped.** The merge SHA does not exist yet, and open PR #160 owns the pin backlog. ## Requirements and spec sections served | R-ID | Spec section | Where | |---|---|---| | R-VULN-01 | Inventory | `platform/vuln-scanner/src/inventory.py` | | R-VULN-02 | Scanning | `src/main.py`, `src/trivy.py` | | R-VULN-03 | Coverage | `src/state.py`, `src/report.py` | | R-VULN-04 | The CVE page | `src/page.py`, sidecar, HTTPRoute (**login blocked, see operator action**) | | R-VULN-05 | Trigger, no human step | scan loop, `.gitea/workflows/cve-triage.yaml` | | R-VULN-06 | Alerting | `platform/vuln-scanner/alerts.yaml`, `src/metrics.py`, promtool tests | | R-VULN-07 | The triage bot | `.gitea/workflows/scripts/cve_triage.py` | | R-VULN-08 | Risk acceptance | `platform/vuln-scanner/acceptances.yaml`, `src/acceptances.py`, L0 check | | R-VULN-09 | Security of the scanner | `rbac.yaml`, `networkpolicy.yaml`, digest pins, `renovate.json5` | Also surfaces R-OBS-02/03, R-C2-02, R-UPD-01/02, R-SEC-02, R-IDP-06. Own Flux tier (`clusters/base/vuln-scanner.yaml`): `wait: false`, `dependsOn: platform-config`, nothing depends on it, so this tier cannot stall the platform. ## Gate Local toolbox gate on `e1adc39`: **rc 3, UNDETERMINED, not green.** ``` [skip] stage spec-immutability: SKIPPED (not run) [skip] stage gitea-settings: SKIPPED (not run) [skip] stage conformance: SKIPPED (not run) [ ok ] approval-policy: green [ ok ] decisions-log: green [ ok ] lint: green [ ok ] l0: green [ ok ] render-golden: green [ ok ] l1: green [fail] Stages that did not run: spec-immutability gitea-settings ``` `conformance` (L2/L3) skipped: no cluster. `approval-policy` is N/A offline and proves nothing locally. L2/L3 tests for this feature are scaffolds that do not run in CI. Renovate validator (`renovate:44.167.0`, the tag the workflow runs): validated successfully. ## Operator action required 1. **Review the sensitive paths.** This PR touches `.gitea/**`, `conformance/**`, `clusters/**` and RBAC, so `approval-policy` is expected to be red in CI until an administrator reviews. That is the correct stopping point. No agent will post `/approve-override`. 2. **Add the Keycloak client. R-VULN-04 is NOT delivered until this is done.** Login fails closed (`invalid_client`) until then; the scanner, metrics and alerts are unaffected. An agent cannot do it: the client secret must be a new key in the existing `platform/keycloak/keycloak-realm-clients.sops.yaml` (needs the age private key), and the live realm needs a one-time `kcadm` client creation. Exact ordered steps are in `platform/vuln-scanner/README.md`, "Operator action required". Those steps are UNVERIFIED. ## Expected after deploy - **VulnCriticalFixAvailable will fire on day one.** That is the backlog becoming visible, as ADR-0027 says. - VulnTriageBotSilent fires until the bot's first successful run at 07:15 UTC or a `workflow_dispatch` run. - VulnScanStale fires during the first full scan cycle, because the last-complete metric is absent until one completes. ## Spec escalations 1. `credential-required` is defined as "the registry refused anonymous access", but the egress contract excludes LAN ranges, so the scanner cannot reach the private registry to be refused. Implemented by naming private registries in `VULN_PRIVATE_REGISTRIES`, short-circuiting before Trivy. 2. The egress contract lists only the Kubernetes API, DNS and public TCP/443, but the SSO sidecar must reach Keycloak. Egress additionally allows the Gateway on 443 and Keycloak. ## Deviations from the recommended shape - Page is server-rendered with no JavaScript rather than inline JS: no `script-src` in the CSP, one escaping path, and L1 tests exercise the real rendered page. - Three pod ports (page, internal, metrics) so Prometheus cannot read `report.json`. - Python 3.14.9-alpine, the version the tests ran on; Trivy 0.76.1, as 0.76.2 was not old enough. ## UNVERIFIED until it runs in production Nothing has been deployed. No runtime behaviour has been observed. - A real Trivy run; every scanner test uses a recorded fixture and a fake runner. - The real Kubernetes API, browser rendering, OIDC login and the `platform-admins` restriction. - In-cluster DNS from a DinD job container, and reachability of the pinned ClusterIP fallback `10.43.210.10`. The workflow has never run; Gitea behaviour is tested only against a fake. - Renovate's `## Pending Approval` row format, because issue #91 has no live approve row today. - `minimumReleaseAge` behaviour on a real update. - Cgroup v2 memory behaviour under the scanner's limits (inferred). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
gitea-admin added 15 commits 2026-11-09 18:41:24 +00:00
feat(vuln): CVE triage bot workflow, script and L1 tests c3c72c47bb 
Scheduled Gitea Actions bot in the Renovate shape: maintains a
counts-only CVE Dashboard issue, derives remediation states from
Renovate's dashboard and PRs, ticks approve rows for exact-matched
images with an unaccepted CRITICAL with a fix, and posts the heartbeat.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
feat(vuln): point the triage bot's fallback at the pinned scanner ClusterIP ec6b39bb91 
Co-Authored-By: Claude Opus 5.5 <[email protected]>
feat(vuln): scanner service and fixture-driven L1 tests 50ec2297ee 
Stdlib-only scanner per docs/17: Kubernetes API inventory, Trivy runner,
persisted state, schema-validated report.json, strict acceptances parser,
server-rendered page, aggregate metrics, internal API, and a scan loop
that stays live while scanning.

L1 tests under conformance/tests/l1-rendering/vuln-scanner-* run the
actual modules against recorded fixtures with a fake Trivy runner.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
feat(vuln): vuln-scanner platform manifests, Flux tier, alerts, L0/L2/L3 tests b85d1e3ed7 
Hand-authored platform/vuln-scanner (non-gating Flux tier, digest-pinned Trivy 0.76.1 and
python 3.14.9-alpine, read-only RBAC, per-pod oauth2-proxy, internal-only cve.<domain>),
five alerts with promtool tests, R-VULN-01..09 registry entries, and vuln-* L0 checks with
mutation self-tests. src/ and the bot workflow are integrated separately.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
feat(vuln): encrypted oauth2-proxy credential for the CVE page d4bea8a78c 
Generated through pipes and encrypted to the repo's public age recipient; values never echoed.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Merge branch 'wip/vuln-scanner' into feat/vuln-visibility 2b9dcef6c7
Merge branch 'wip/vuln-bot' into feat/vuln-visibility b265d08278
Merge branch 'wip/vuln-platform' into feat/vuln-visibility b1a6fe6503
fix(vuln): no-per-cve-label check also matches escaped-quote label forms d06dff9a19 
Co-Authored-By: Claude Opus 5.5 <[email protected]>
docs(vuln): list the CVE triage workflow in the workflows README 4c82c64b2f 
Co-Authored-By: Claude Opus 5.5 <[email protected]>
fix(vuln): metric label extractor handles escaped quotes in f-strings 1763e27859 
Co-Authored-By: Claude Opus 5.5 <[email protected]>
fix(vuln): survive OOM kills, isolate bad upstream strings, label findings per line d8b25a67fe 
Review findings on the integrated scanner:
- An OOM kill no longer wedges the loop: inFlight is persisted before each scan, and on restart
  that image is marked failed and sits out full runs for 20h.
- One bad upstream string can no longer freeze the report: safe_url requires an exact lowercase
  scheme, acceptance reasons are bounded, /triage urls use the same rule, and report.build
  repairs or drops any entry the schema would refuse. lastCompleteRunAt advances only after a
  report is actually published.
- The private registry reports credential-required, not failed: VULN_PRIVATE_REGISTRIES names
  it, and Trivy is not invoked.
- loadError clears after a full run against a good inventory.
- Every image under platform/vuln-scanner stays out of Renovate auto-merge.
- The page labels status and target state per line, and mixed acceptance reads as mixed.

The Keycloak client is still an operator action: its key must go in an existing encrypted file
that an agent must not modify. platform/vuln-scanner/README.md has the exact steps.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Merge branch 'wip/vuln-platform' into feat/vuln-visibility 1ff527f546
Merge branch 'wip/vuln-fixes' into feat/vuln-visibility 4ca0ebfac6
docs(vuln): handoff for automated CVE visibility
gate / gate (pull_request) Successful in 1m5s
Details
gate / conformance (pull_request) Skipped
Details
e1adc3963c 
Co-Authored-By: Claude Opus 5.5 <[email protected]>
jakman85 approved these changes 2026-11-09 22:17:47 +00:00
Dismissed
gitea-admin added 1 commit 2026-11-09 22:29:37 +00:00
Added required keycloak config
gate / gate (pull_request) Failing after 1m0s
Details
gate / conformance (pull_request) Skipped
Details
1ace03569c
gitea-admin dismissed jakman85's review 2026-11-09 22:29:37 +00:00
Reason:

New commits pushed, approval review dismissed automatically according to repository settings

gitea-admin added 1 commit 2026-11-09 22:31:27 +00:00
Updated SPEC_PIN
gate / gate (pull_request) Failing after 59s
Details
gate / conformance (pull_request) Skipped
Details
fd462a4ede
jakman85 approved these changes 2026-11-09 22:33:46 +00:00
Dismissed
gitea-admin added 1 commit 2026-11-09 22:40:25 +00:00
Whitespace linting fix
gate / gate (pull_request) Successful in 58s
Details
gate / conformance (pull_request) Skipped
Details
6680f361a1
gitea-admin dismissed jakman85's review 2026-11-09 22:40:25 +00:00
Reason:

New commits pushed, approval review dismissed automatically according to repository settings

jakman85 approved these changes 2026-11-09 22:40:45 +00:00
jakman85 merged commit 190cd6b81a into main 2026-11-09 22:41:50 +00:00
Sign in to join this conversation.
Reviewers
No Reviewers
jakman85
Labels
Clear labels
agent
automated-update
blast-radius
breaking
ci-tooling
dependencies
needs-review
own-image
range-widen
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
gitea-admin jakman85 (Keith Jackson) pm-agent renovate-bot
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Home-Network/home-network-k3s-impl#165