CTIPilot
AI-generated · no human review · verify critical claims against the linked source. how it works →

2026-11-20T0356Z-intel

One pipeline fire, in full · intel run of 2026-11-20 · sub-agent allocation and telemetry, per-iteration verification verdicts and findings, source-list edits, coverage gaps, bridge invocations, and the run's own verification & coverage notes: what was published, what was dropped at the borderline or judged not relevant (and why), single-source carve-outs, and contradictions. Rendered from runs/2026-11-20/2026-11-20T0356Z-intel.md.

Run telemetry

2026-11-20T0356Z-intel intel prompt v4.25 publish ok
4h 06m duration 5 published 6 updates
Claude Sonnet 5.5 (claude-sonnet-5-5) main agent
S1 Claude Sonnet 5.5 (claude-sonnet-5-5)
Items returned
16
Duration
38m 03s
Tool calls
7 WebFetch13 WebSearch150 bridge
Cited sources
6 of 26 in slice
S2 Claude Sonnet 5.5 (claude-sonnet-5-5)
Items returned
6
Duration
22m 42s
Tool calls
6 WebFetch34 WebSearch70 bridge
Cited sources
2 of 19 in slice
S3 Claude Sonnet 5.5 (claude-sonnet-5-5)
Items returned
6
Duration
22m 06s
Tool calls
6 WebFetch8 WebSearch70 bridge
Cited sources
2 of 14 in slice
S4 Claude Sonnet 5.5 (claude-sonnet-5-5)
Items returned
7
Duration
20m 29s
Tool calls
4 WebFetch22 WebSearch118 bridge
Cited sources
4 of 8 in slice

Verification

#1 NEEDS_FIXES · Sonnet 5.5 · t=18 e=8 a=6 #2 NEEDS_FIXES · Sonnet 5.5 · t=12 e=5 a=4 #3 NEEDS_FIXES · Sonnet 5.5 · t=4 e=1 a=5 #4 NEEDS_FIXES · Sonnet 5.5 · t=2 e=1 a=4 #5 NEEDS_FIXES · Sonnet 5.5 · t=7 e=1 a=4 #6 NEEDS_FIXES · Sonnet 5.5 · t=2 e=0 a=6

1 entry dropped by verification this run (recorded in the verification & coverage notes).

Deep dive

·

Entries this run published (5) and updated (6)

Sources changed (this run)

Edits this run made to sources/sources.json · promotions, demotions, new candidates, and fetch-method / category / reliability / url corrections (the run record's sources_changed[]). Paginated; 10 per page.

3 added candidate · 1 promoted to active · 1 notes · 1 last_successful_fetch · 1 content_scope · 1 url.

SourceChangeFrom → ToReason
broadcom-vmware-security-advisoriesadded candidate· → ·no VMware PSIRT record existed although vCenter is a hypervisor management plane; VMSA pages and the Q&A READMEs read with extract
reversinglabs-blogadded candidate· → ·original malware research with detection guidance; feed verified
onlinereports-chadded candidate· → ·Basel regional outlet that carried the Basel-Landschaft statements while baselland.ch refused every transport
unit42-timely-threat-intel, nginx-security-advisoriespromoted to active· → ·cited by published entries from at least three distinct runs (sources.promotion_due)
ransomware-live, bs-ch-medienmitteilungen, tio-ch-attacco-informatico, borncity, divd-csirtnotes· → ·working recipes found by S2 and S4
advisories-ncsc-nl, anssi-fr, bsi-de, enisa-euvd, ncsc-ch-security-hub, wordfence, heise-sec, inside-it-ch, opensourcemalware, hunt-io, borncity, divd-csirt, next-ink, tio-ch-attacco-informaticolast_successful_fetch· → ·fetched and used by an entry or a changelog record published this run; counters reset
typo3-security-advisories, polizeiticker-chcontent_scope· → ·source_health UNSOLVED list: the typo3 feed titles are CWE class names and polizeiticker relays every canton, so neither carries the generic security vocabulary; readable, dated and current is the check (general-news scope)
broadcom-vmware-security-advisoriesurl· → ·source_health UNSOLVED list: github.com is refused by the container egress; the record now points at the VMSA-2026-0009 advisory page that extract reads

Coverage gaps (this run)

Sources this run's brief needed that returned no usable content via any documented recipe. Bridge-recovered or quiet-day sources do NOT appear here. (Distinct from the independent source-accessibility probe at the foot of this section, which probes all active sources regardless of what any run needed.)

Source (uncovered)URL triedMethod chainStatus / classWhat the agent did instead
baselland-chhttps://www.baselland.ch/politik-und-behorden/regierungsrat/medienmitteilungenbridge:extract → bridge:url → webfetch → bridge:jina403 transport-403
canton site answers 403 on extract, url and WebFetch; the reader pool is dead
the chancellery statements were read through Keystone-SDA relays (nau, OnlineReports, Bajour, Prime News); the incident stays held on the backlog
usi-chhttps://www.usi.ch/it/data-breachbridge:extract → webfetch → bridge:jina403 transport-403
USI notice page answers 403 on every transport
the communique was read as reproduced by Corriere del Ticino and laRegione, which agree
f5-k-articleshttps://my.f5.com/manage/s/article/K000164058bridge:extract → bridge:url-direct → webfetch → bridge:jina200 js-shell
myF5 single-page application: extract, url --direct and WebFetch return an empty shell
F5 text read through its CVE Program records and the CERT-FR, NCSC-NL and BSI restatements; the BIG-IP entry was then dropped after verification because the per-flaw conditions could not be cited to a readable page
perforce-portalhttps://portal.perforce.com/bridge:extract → webfetch200 js-shell
Salesforce shell with no advisory text
Perforce statements read through ENISA EUVD, NCSC.ch, SecurityWeek and the OpenVox advisories
belden-psirthttps://www.belden.com/securitybridge:extract → webfetch → bridge:jina403 transport-403
listing 403 on every transport, reader pool dead
none new this fire; advisory PDFs on assets.belden.com remain the route
lab52https://lab52.io/blog/bridge:feed → bridge:extract → webfetch → bridge:jinan/a connection-reset
the feed resets the connection and the listing is undated
one outside-slice try; nothing newer visible

Bridge invocations (this run)

7 bridge calls this run · these are successful bridge fetches (separate from "Coverage gaps" above).

7 other
  • tools/fetch_source.py cisa-kev and tools/kev_window_diff.py ×1
  • tools/fetch_source.py extract ×1
  • tools/fetch_source.py url --direct ×1
  • tools/fetch_source.py ncsc-csh post ×1
  • tools/fetch_source.py pdf ×1
  • WebFetch with the outbound-links template ×1
  • tools/fetch_source.py jina (last rung) ×1

Verification findings · all iterations

Per-iteration finding detail. Each table is one verifier pass · what was flagged, how the main agent remediated it, and the outcome. Walking the tables top-to-bottom shows the verifier's debugging trail across iterations.

Iteration #1 NEEDS_FIXES · 32 findings (truth=18, editorial=8, advisory=6) · Claude Sonnet 5.5 · 26m 50s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F4
hallucinated-fact
·
GHSA-8mjj says 'Fatal error backtraces are enabled by default in PHP 8.5. PHP's built-in default also includes arguments, but php.ini-production sets zend.exception_ignore_args=1 and prevents this atthandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F14
?
·
No cited source says CVE-2026-129289 is the only such flaw. GHSA-qc49 itself requires 'an application path that opens php://input without reading it fully'; GHSA-gr4f-wgx5-jpfv (CVE-2026-129496) says handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F14
?
·
The branch changelogs differ: 8.5.13 lists 18 CVEs, 8.4.28 lists 14, 8.3.37 lists 13, 8.2.36 lists 12 (counted in the raw ChangeLog-8.php HTML, id anchors 8.5.13, 8.4.28, 8.3.37, 8.2.36); only 8.5.13 handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
The record's fields are [summary, actions, cves, sources, body]; immediate_action was left unchanged. The update itself says PHP calls CVE-2026-129481 'an incomplete fix for CVE-2026-109619' and actiohandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
GHSA-658j: 'Check for existing configuration that explicitly enables it, because an explicit setting overrides the new default ... If session upload progress is explicitly enabled ... all readers and handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
Elastic's post says only 'The ECK operator does not use Java and is therefore not affected' and 'Elasticsearch clusters that ECK manages are affected in the same way as other self-managed Elasticsearchandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F3
claim-not-supported
·
The case page carries the AI-agents-after-root text but not the credentials sentence; that is on https://csirt.divd.nl/cases/DIVD-2026-00014/overview_data_investigation/ ('The attacker was mainly intehandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F3
claim-not-supported
·
GHSA-34m4 covers only the one-byte overflow; the double free is GHSA-7x2q-68qv-6p8x and the regexp integer overflow GHSA-3p8q-wmhx-vm2w (both Critical per their GHSA labels), the ruby-lang advisory cahandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
MISA-2026-0013 names only 'the Platform component of OpenScape 4000 and OpenScape 4000 Manager'; 'web services', 'numbering and trunk configuration' and 'Linux base' appear in neither Mitel's text norhandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
F5's CNA records say 'undisclosed requests can cause ... TMM to terminate and may lead to remote code execution (RCE)'; NCSC-NL says 'Sommige van deze kwetsbaarheden' (some of them) can be used for cohandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F3
claim-not-supported
·
BleepingComputer says 'part of the same intrusion campaign' and nothing on victim sector or country; the GTIG page thanks Broadcom for collaboration and does not say Mandiant reported the flaw (Broadchandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F14
?
·
GTIG: 'ARMORY CASTLE disabled the SSH service via the VAMI API immediately after extracting the vCenter master symmetric key and PostgreSQL database credentials.' No duration is given. Use 'shortly afhandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
GTIG describes extracting specific files (NTDS, SYSTEM and SECURITY hives, the DNS directory) from the guest VMDKs with vmfs_browser.py, then archiving o.tar.gz; the body says 'read Windows guest virthandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
USI's communiqué (as reproduced by Corriere del Ticino and laRegione) says third parties had access to data on the server via a software vulnerability; it does not say the server was internet-facing ohandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
NCSC.ch post 13205 has no detection content; GHSA-qc49 says only 'a malformed multipart/form-data body that makes PHP's multipart parser stop at a predictable position', not an unterminated boundary. handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F3
claim-not-supported
·
The advisory's sidebar shows 'Patched versions None' for 8.1.20 to 8.1.34 but never says 8.1 is end of life. The 2026-10-21 php.net archive item says 'PHP 8.1 and older no longer receive updates from handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F4
hallucinated-fact
·
The case page states the 19 h gap and 'we see plenty of room for improvement' but names no lesson; the overview says the shared-mailbox OAuth tokens 'were obtained by the attacker and used to send teshandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F3
claim-not-supported
·
EUVD-2026-116493 shows baseScore 9.0 only. The 9.5 appears as the GitHub score badge on GHSA-q4hm-33rc-572g, whose own text says 'CVSS 4.0: 9.0 (Critical)'. Drop the 9.5 or cite the badge explicitly.handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F9
surface-contradiction
·
The Commission transcript says 'the incident was detected on the 22nd of October and it was confirmed ... on the 26th of October'; POLITICO (published 2026-11-19) says 'CERT-EU, detected earlier this handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F9
surface-contradiction
·
VMSA-2026-0009.1 response matrix lists 13.x.x.x and '12.3.x, 12.4.x, 12.5.x'; the Q&A (which reflects the initial advisory) says 'CVE-2026-123894 affects VMware Tools for Windows 13.x, 12.x, and 11.x handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F5
missing-citation
·
The sources[] are CERT-FR, NCSC-NL and BSI, none of which carries the per-CVE prerequisites (HTTP profile for PSC::ip_address and MR::message route, APM agents), and sources[0] is a national-CERT resthandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F8
needs-more-research
·
F5 CNA record: 'When BIG-IP APM is provisioned and a per-request policy that includes the Client Information agent and Endpoint Security agents is applied to a virtual server'. Any per-request policy handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F16
?
·
The critical bar lists 'patches >=1 week old without new exploitation' as a disqualifier; the 2026-11-20 delta is a regular release and follow-up fixes, no new exploitation. A non-internal update floahandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F10
missed-angle
·
A confirmed multi-source incident at a Swiss cantonal administration (constituency core) with a public-warning contingency is held while USI, with the same absence of product and actor, ships as routihandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F18
?
·
The same checks are in the body Detection line; as an action it is a concrete compromise-check task but duplicates the body and carries the unsourced 'within minutes'. Keep it as one task naming the Shandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F7
drop
·
The section adds no ARTEX behaviour, vector or detection (the run record says so: 'incident floor'), the Commission does not name the tool and the link rests on one anonymous official, yet the record handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
The Detection line (main analysis) does not contain the three checks; the section describes an edit to it that was not made, and the supersession sweep asks for the line itself to be rewritten. Fold thandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
'reader' is fetch-transport vocabulary and 'could not be read' is process narration in a reader-facing summary; the rule is two sentences of provenance in sourcing_note. Say 'F5's bulletin is a login-handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
In the Veria page the comma follows </time></strong> directly ('25 November 2026, or later'); the space is a normalisation artefact, so the quote is not a contiguous rendered substring. The USI quote handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
With no product, vector detail or actor, the three labelled lines are generic advice derived rather than sourced, on an entry that should be two sentences at most by the incident floor; trim to the fihandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
The newly described certificate-name impersonation (node impersonation, catalog theft, CA administration) has no technique id in techniques[]; the record's fields do not name techniques.handled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml
F11
editorial-advisory
·
The PHP update replaced sentences rather than cutting them, so the analysis is not brought toward the bound by the record that touched it; the audit sweep is the place to compact, but a record that edhandled before iteration 2: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter2.yaml

Iteration #2 NEEDS_FIXES · 21 findings (truth=12, editorial=5, advisory=4) · Claude Sonnet 5.5 · 24m 16s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F14
?
·
iteration-1 finding #16 was remediated in the body ('The most direct is CVE-2026-129289') and the backlog row but NOT in the frontmatter summary, which still carries the unsupported quantifier. GHSA-ghandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
the run replaced the October citation (The Hacker News) for 'no exploitation is reported' with the GHSA, which states nothing about exploitation in the wild (its only exploitation-related text is 'Exphandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
(low confidence) the cited GHSA says only 'introduced ... first released in PHP 8.1.20 and 8.2.7' and 'Upgrade to the fixed release for your PHP branch when it becomes available'; it never says 8.1 gehandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F4
hallucinated-fact
·
(low confidence) the owning advisories say authentication depends on the application path, so a flat pre-auth record overstates; NCSC.ch's generic 'unauthenticated remote attackers' covers the set onlhandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
(low confidence) the overview attributes the 207 email addresses, two phone numbers and seven profile pictures to 'Our Jira and Confluence environment was compromised', not to the Zammad host whose blhandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
(low confidence) POLITICO: 'Disruption has lasted over 10 days. Sysper ... and MyPMO ... were restored earlier this week'; the 10 days apply to the disruption, not to those two systems being offline thandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
(low confidence) the cited page carries only the second half ('It also did not establish whether ARTEX helped them break in or was used only to look for weaknesses'); that the Commission has not confihandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F4
hallucinated-fact
·
(low confidence) Veria's page says 'Veria Labs found this vulnerability independently and was the first to demonstrate remote code execution' and 'The Ruby security advisory also credits Stephan Lipp handled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F4
hallucinated-fact
·
(low confidence) the advisory names only 'adjacent network access' (MTLVULN-1793) and 'unauthenticated attacker'; 'user, partner or internet networks' appears nowhere in it, yet the clause carries Mithandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F3
claim-not-supported
·
(low confidence) GTIG places vpxuser on the ESXi Host Client and only root on SSH, and does not state the source host of either login; the clause makes both SSH logins and names the source.handled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F13
?
·
(low confidence) NCSC.ch lists the Reddit thread under 'Additional Sources' as 'Community report of in-the-wild exploitation' next to the CISA KEV entry and never says its label rests on it. THN quotehandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F4
hallucinated-fact
·
the record on entries/2026-10-21/php-critical-release-... is `type: improvement` after iteration 1; the Backlog bullet and the struck row in state/coverage_backlog.md still say `update`, and the noteshandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F8
needs-more-research
·
(low confidence) the rewrite fixed the 'non-default' error but now states exposure by configuration only; the advisory's application prerequisites (reachable __call/__callStatic path, request-controllhandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F8
needs-more-research
·
(low confidence) the interim firewall action omits GTIG's own testing and change-control caveat for the iptables rules it comes from; GTIG describes port 2012 as 'the control interface RPC for vCenterhandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F9
surface-contradiction
·
the advisory (revised 2026-11-20 'to clarify the impacted VMware Tools versions') restricts 12.x to 12.3 to 12.5, the Q&A says 12.x and 11.x; the entry writes '12.x' as if both agreed and attributes ohandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F5
missing-citation
·
the sentence names NCSC.ch's measures but carries no inline link; NCSC.ch lists them ('Temporary: Disable certificate auto-signing and enforce strong CSR validation. Audit issued certificates for susphandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F5
missing-citation
·
uncited; the case page says 'The forensic investigation of the breach has been completed' and records 'Updated the casefile ... 18 Nov 2026' but states no completion date, so 'closed ... on 2026-11-18handled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F11
editorial-advisory
·
the precondition is an endpoint or framework that opens php://input without consuming it, not one that reads it early; also a stray comma before the citation ('are not affected by the demonstrated atthandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F11
editorial-advisory
·
(low confidence) the section reports a genuinely new development (new fixed builds, three CVEs, a changed update target and interim advice); typing it `improvement` avoids the re-float rather than deshandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F11
editorial-advisory
·
the improvement record covers `body` but neither the section nor the summary mention that two ESET-supported statements (the attribution basis, which ESET's page states, and the CerealHermit variant) handled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml
F11
editorial-advisory
·
iteration-1 #29 was answered by the summary note only; the Zammad entry's convention puts the marker in the quote itself. The takeaway dates the vulnerable software from 2026-11-09, but USI gives onlyhandled before iteration 3: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter3.yaml

Iteration #3 NEEDS_FIXES · 10 findings (truth=4, editorial=1, advisory=5) · Claude Sonnet 5.5 · 22m 45s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F3
claim-not-supported
·
The live Veria page (https://verialabs.com/ruby2shell/, 'Updated 20 November 2026 at 06:00 UTC') now lists GitLab as 'GitLab | 19.6.0 and earlier | 19.6.1, 19.5.8 | Patched'; only GitHub Enterprise Sehandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F4
hallucinated-fact
·
Zammad released 7.2.3 on 2026-11-11 (https://zammad.com/en/product/releases/7-2-3, title 'Zammad 7.2.3: Critical Security Update') and published 43 GitHub advisories the same day, every one with vulnehandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F3
claim-not-supported
·
(low confidence) GHSA-jm4m-qfqx-967m says 'execute commands with the privileges of the FPM master, which commonly runs as root' and 'If the master runs as an unprivileged user, successful exploitationhandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F14
?
·
(low confidence) GHSA-8mjj-2r7p-4hg2 says php.ini-production sets zend.exception_ignore_args=1 and prevents the attack path, and also 'On PHP 8.5 and later, setting fatal_error_backtraces=0 also prevehandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F10
missed-angle
·
No entry or changelog record in the store covers the 2026-11-11 Zammad release; the store's only Zammad entries are the 2026-10-02 exploited-flaw entry (which this run updated) and a June BSI entry. Zhandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F11
editorial-advisory
·
The body section was corrected to 'completed its forensic investigation and updated its case file on 2026-11-18' but the record summary still says 'closed'. DIVD's case page says 'The forensic investihandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F11
editorial-advisory
·
Typography: a full stop closes the clause before the citation parenthesis, so the ledger reads the citation as detached. Also `entities: []` although the registry carries product:mitel-openscape-4000 handled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F11
editorial-advisory
·
GHSA-fm2m-fjf5-v3cp says 'Authentication requirements depend on how the application exposes the affected session operations' and GHSA-g3vh-838c-65v7 says 'Authentication requirements depend on the apphandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F11
editorial-advisory
·
The Hacker News and SecurityWeek list Puppet Core (9.0.0 before 9.1.3, 8.x before 8.22.2) as affected by CVE-2026-115731 and the entry's action and affected_products already name Puppet Core, but the handled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml
F11
editorial-advisory
·
A routine incident with no named product, vector detail or actor is meant to be two sentences at most; the body is one three-sentence paragraph plus the takeaway. The Swiss public-sector nexus justifihandled before iteration 4: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter4.yaml

Iteration #4 NEEDS_FIXES · 7 findings (truth=2, editorial=1, advisory=4) · Claude Sonnet 5.5 · 19m 10s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F3
claim-not-supported
·
(low confidence) Frontmatter summary says 'Cisco lists 13 products as affected with no fixed release yet' and the body says 'Cisco lists 13 products, Catalyst Center, Secure Network Analytics and Secuhandled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F4
hallucinated-fact
·
(low confidence) Frontmatter tag poc-public overstates the body: the body says Ruby is 'aware of proofs of concept', Veria 'plans to publish a proof of concept ... on 25 November 2026' and NCSC.ch 'Prhandled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F10
missed-angle
·
(low confidence) The VulnCheck page the entry cites for its 2026-10-26 build list also says 'As of November 10, the team has added RCE support for the following older PHP versions ... PHP 7.4.10-7.4.3handled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F11
editorial-advisory
·
Advisory: headline and summary still present 7.2.2 as the newest fix; the 7.2.3 supersession (critical security update of 2026-11-11) lives only in actions, cves[].fixed, the takeaway and the new secthandled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F11
editorial-advisory
·
Advisory: vCenter lists VMware SDDC Manager and VMware Tools in affected_products but not product:vmware-sddc-manager / product:vmware-tools in entities (both keys were registered this run); the Java4handled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F11
editorial-advisory
·
Advisory (very low): the NCSC.ch post words the third Critical flaw as 'A heap overflow from a regex size overflow during compilation'; 'integer overflow' is the Ruby announcement's wording (GHSA-3p8qhandled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml
F11
editorial-advisory
·
Advisory (very low): the follow-up CVEs the entry names (CVE-2026-129481, CVE-2026-129182) do not apply to 8.2 (GHSA-g3vh: PHP 8.2 does not contain the affected field; GHSA-2fjv: 8.5 only), so '8.2.36handled before iteration 5: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter5.yaml

Iteration #5 NEEDS_FIXES · 12 findings (truth=7, editorial=1, advisory=4) · Claude Sonnet 5.5 · 27m 09s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F3
claim-not-supported
·
cves[0].fixed still reads 'Cisco Catalyst Center, Secure Network Analytics, Secure Workload and ten more Cisco products: fixed release to be determined'. Cisco revision 1.2 (2026-NOV-19) lists 22 affehandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F3
claim-not-supported
·
(low confidence) Section says 'as of 2026-11-18 Red Hat and Ubuntu had released no fix and Debian had fixed only its unstable branch, so hosts on those distributions had no stable vendor build yet'. Thandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F3
claim-not-supported
·
(low confidence) The cited NCSC.ch post lists a Moderate regexp flaw (GHSA-pvgc-g7gj-pv6h) that needs only 'A crafted subject string matched against a fixed pattern that uses a case-insensitive backrehandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F3
claim-not-supported
·
(low confidence) The cited Cisco advisory is now revision 1.11 (2026-NOV-06, Final): 'Updated Ultra-Reliable Wireless Backhaul (URWB) and Unified Industrial Wireless Software as being available'; the handled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F3
claim-not-supported
·
(low confidence) The cited FG-IR-26-186 timeline ends '2026-11-05: FortiSOAR and FortiSOAR PaaS added to impacted product list'; 'Products Under Investigation' lists only FortiDDoS, and FortiSOAR rowshandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F3
claim-not-supported
·
(low confidence) GreyNoise's live tracker (as of 20 November) lists the first shell at '2026-10-27 16:25:05' (apache2 -> apache2, a miner command) and says first successful exploitation came '1 day lahandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F4
hallucinated-fact
·
(low confidence) Neither quote is a substring of its page as it reads on 2026-11-20. GreyNoise tracker now says 'An attacker exploited a vulnerable GreyNoise sensor: apache2 spawned apache2, which ranhandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F10
missed-angle
·
(low confidence) The ENISA EUVD record the entry cites lists this public repository among its references: an independent root-cause write-up with ASAN crash reproduction and local, no-ASLR weaponizatihandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F11
editorial-advisory
·
Carried over from iteration 4: headline and summary still present 7.2.2 as the newest fix; 7.2.3 (critical security update, 2026-11-11) lives only in actions, cves[].fixed, the takeaway and the newesthandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F11
editorial-advisory
·
Registry keys exist but are not linked: Puppet entry (entities: [] although product:perforce-puppet-enterprise, product:puppet-core, product:openvox-server are registered, two of them 'added' this runhandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F11
editorial-advisory
·
FIRST value was 0.00366 on 2026-11-17 but the deep read recorded 0.00691 on 2026-11-19 ('stale'); the record names `cves` in fields, so the refresh was cheap. Not a truth defect (a dated snapshot).handled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml
F11
editorial-advisory
·
(a) The 2026-11-19 releases are a new development (a new fixed build, a changed update target, CVE-2026-115560 closed) typed as an improvement so the critical entry does not re-float; a defensible chohandled before iteration 6: fixed, declined with reason or reverted, per work/2026-11-20T0356Z-intel/deltas.iter6.yaml

Iteration #6 NEEDS_FIXES cap-breach · 8 findings (truth=2, editorial=0, advisory=6) · Claude Sonnet 5.5 · 23m 55s

F-codeSectionItem · URL/quoteVerifier summaryRemediation · outcome
F3
claim-not-supported
·
The iteration-5 remediation dropped the hedge together with the stale count. In a `fixed` field the Cisco products now read as having fixed builds, but Cisco's rev 1.2 (2026-NOV-19) table says Catalysfixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F3
claim-not-supported
·
(low confidence) Adjacency: the clause is terminated by the two GHSA pages, and neither says anything is 'left open' in October (GHSA-658j: 'This is a configuration-dependent deserialization issue, sefixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
Garbled sentence in a reader-facing field: 'supersedes it installations' is an edit residue (a clause lost between 'supersedes it' and 'installations'). The facts hold (Zammad 7.2.3 page: 'a security fixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
Both statements are dated to the previous fire and were true then. Cisco rev 1.2 (2026-NOV-19) now lists IOS XE Software and UCS Central Software as 'Not affected'; Tenable's page now says 'As of Novefixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
The note holds only for the top-level `references` field. The same JSON response carries a nested record (GHSA-8452-w857-wpmw, alias of CVE-2026-132825) whose `references` include https://github.com/efixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
GTIG's text is 'ARMORY CASTLE exploited this vulnerability in early September 2026 using BONESAW, according to forensic evidence' and 'BONESAW was executed in early September 2026' in one intrusion; 'fixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
GHSA-8mjj also says 'The demonstrated code execution chain also requires knowledge of usable memory addresses despite address space randomisation' and 'Authentication requirements depend on that path'fixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit
F11
editorial-advisory
·
One-day drifts that match the pages' modified dates, so probably intended as the as-of date; the rule asks for the publication date. Low value; fix only if the entries are touched again.fixed after the pass where it was a truth or editorial finding (Java4Shell fixed field qualifier restored, PHP clause cited to watchTowr, Zammad summary clause repaired); advisory items left to the quality audit

Verification & coverage notes

The run record's narrative body, verbatim. This is where the run accounts for its own judgement calls: every borderline drop and judged-not-relevant item with its reason, dedup decisions, single-source items and their carve-outs, contradictions, and per-source coverage gaps, so nothing the run considered disappears silently.

Verification & coverage notesrun record body

2026-11-20T0356Z-intel · Sonnet 5.5 · window 26 h · 5 entries published

Verification & coverage notes

Window 26 h (gap 24.0 h to 2026-11-19T0355Z-intel, class Standard). Five new entries and six changelog records, no deep dive, no critical. The will-publish set was 14 items, over the ~8 a main-agent deep read should carry, so two scoped cti-research follow-ups read the primaries and returned claim tables whose passages were grep -F checked against saved bodies: DR1 (five vulnerability items, 2026-11-20T04:40:37Z to 05:03:10Z) and DR2 (eight incident and update items, 04:40:37Z to 05:05:59Z); the main agent composed every entry and record from those tables and the findings files. Stop-hook commit nudges arrived mid-run and were not acted on; the commit follows the gate and the verifier loop.

  • KEV sweep: kev_window_diff.py found no CISA KEV addition with dateAdded inside the window (catalog 2026.11.18) and no ransomware-flag row; S1 read the feed too and found the same. No row needed a disposition.
  • Backlog: the PHP row is published (2026-11-20/php-8-5-13-release-fpm-request-smuggling-session-flaws, notable, ; the October pre-announcement entry's "left unfixed" statements for CVE-2026-115560 are now stale and its update was reverted, see below). The Basel-Landschaft row stays held: S2 and S4 found only a class-level statement (the chancellery says the attacks are intrusion attempts, not DDoS flooding) and no vector, behaviour or data taken, no leak-site claim and nothing from BACS; expiry 2026-12-03 unchanged.
  • Priority calls: the vCenter authentication bypass (CVE-2026-123890) is high, not critical: GTIG documents one targeted intrusion that reached vCenter from an internal foothold, Broadcom does not state exploitation, and there is no public proof of concept, scanning or KEV listing, so the critical bar (mass exploitation imminent or an ongoing campaign) is not met on today's facts. Ruby is high because the discoverer announces a public Rails exploit for 2026-11-25 and Rails is exposed by default; PHP is notable (most flaws need non-default conditions, none is exploited), and the October PHP pre-announcement entry was NOT changed (see below).
  • S1 corrections by DR1: Mitel OpenScape 4000 has a workaround (Security Mode restricted access); vCenter exploitation is stated by GTIG; Ruby gained a dated exploit announcement and a product list from Veria Labs (single source).
  • European Commission: S2 and S4 both returned a new-incident item; it went in as an update on the ARTEX entry because the incident gives no vector, no actor and no behaviour of its own (incident floor) and ARTEX is named only by an anonymous official.
  • borderline-drop: F5 BIG-IP K000164058, six unauthenticated TMM stack overflows (S1, composed as a notable entry, dropped after verifier iteration 1); each needs specific iRule commands or an APM policy, F5's bulletin and K-articles are unreadable, the per-flaw conditions rest on CVE Program records the entry cannot cite, and no exploitation is reported; the previous fire dropped the same family. Re-open if F5 or a researcher reports exploitation.
  • Publica (S2, S4): not recorded. The entry already says test or migration data was held at the supplier; Publica's 2026-11-19 statement of the same origin and the external-review mandate add governance only.
  • borderline-drop: WordPress 7.1.5, primary 33 h old, headline flaw needs an Administrator to open a crafted link (Wordfence), auto-updates deliver it, no exploitation
  • borderline-drop: Splunk SVD-2026-1103; management-plane bypass needs reach to splunkd 8089 and an empty pass4SymmKey, no exploitation, regular patch cycle
  • borderline-drop: Dell PowerStore DSA-2026-479, NAS interfaces normally internal, no exploitation, regular patch cycle
  • borderline-drop: IBM Guardium DP 12.2, internal collector interfaces, no exploitation
  • borderline-drop: JetBrains TeamCity CVE-2026-132891, trigger conditions unestablished (vendor page unreadable), single ENISA record, no exploitation
  • borderline-drop: WatchGuard Dimension CVE-2026-86130, log server, no exploitation
  • borderline-drop: cPanel WP Toolkit proxy, single BSI relay, no CVE or conditions, vendor page unreadable
  • borderline-drop: Frappe GHSA-vwqr-4mjv-ffrm; exploited ITW but needs an authenticated portal user, email spoofing and attachment read only, no constituency nexus
  • borderline-drop: BSI Lagebericht 2026 (annual report), Germany-specific statistics, no Swiss figures, awareness only
  • borderline-drop: Canton Zug M365 policy, relaxation of an internal rule, no obligation change for other defenders
  • borderline-drop: nginx2root Swiss exposure counts (Risikomonitor via Netzwoche), banner-based, no primary
  • borderline-drop: depthfirst Narrow Grants, Wide Engines, patched CVEs on regular cycle, MySQL needs high privileges, Oracle needs Create Session, pre-login chain not reproduced, hardening advice generic
  • borderline-drop: React2STEAL, single source, attacker pipeline on GitHub, no defender decision beyond known React2Shell patching
  • borderline-drop: Phantom53/HavoX, victims US individuals and non-constituency regions, ClickFix chain commonplace
  • borderline-drop: Check Point State of Ransomware Q3 2026, outside window, statistics, no takeaway
  • borderline-drop: N-able Mail Assure intrusion, no vector, no data, no Swiss link
  • borderline-drop: Transport for London photocard breach, UK, no vector, no Swiss link
  • Also dropped by S4 (16 borderline_drops lines in findings.S4.yaml): Infoblox (repeat), Softway Medical, CNIL SAN-2026-016, Neuried, Contabo, DomainFactory, Wellington IT, the NVIDIA/Gentlemen claim, Lithuanian customs, IIHF and Reka, each without a constituency nexus or a transferable behaviour.
  • Not returned by S3 but flagged as possibly missed by earlier fires: Kaspersky's Operation ShadowCloud and AuditTeam OpsAgent posts (2026-11-17/18) and Talos UAT-12643 (2026-11-18), all outside the window; left for the audit's re-sweep.
  • Single-source: USI Lugano (single-source-victim, the victim's communique read through two near-verbatim relays); Mitel OpenScape 4000 (vendor advisory; CERT-FR and BSI restate it).
  • Rotation: 100 standard or candidate sources were excluded from the slices because the previous two fires attempted them (runs.recent_attempts); S1 26, S2 19, S3 14 and S4 8 slice records were all attempted, and belden-psirt and sygnia were extra probes recorded as outside-slice checks, not as attempts.
  • Coverage gaps: baselland-ch (403 on every transport); usi-ch (403, relays used); f5-k-articles (JavaScript shell, CNA records used); perforce-portal (shell, EUVD and press used); belden-psirt (403, reader pool dead); lab52 (listing undated, feed resets); golem article bodies (consent wall, titles only); swisspost-cybersecurity (no dates); senthorus-ch (weekly roundups only).
  • Reader pool: the jina keys answered HTTP 402 almost throughout; no source needed it.
  • Ledger note: S3 rewrote url-liveness.tsv twice (removing lines it had logged in error) while other agents appended, and DR2 reports two portal.perforce.com lines that record a 200 for JavaScript shells; neither is cited.
  • Home-region watch item: Canton Basel-Landschaft (held, expiry 2026-12-03). Re-check search for the next fire: Kanton Basel-Landschaft Cyberangriff Verwaltung Server neu installiert plus the BACS and NCSC.ch hub pages; publish when a source names a vector, behaviour or data taken.
  • Verifier iteration 1 findings declined with reason: the European Commission record stays an update on the ARTEX entry (a new government victim is a development PD-8 names; the section now states the date disagreement); CVE-2026-115732 gets no techniques[] id because no source names a technique the existing mapping lacks; the old PHP entry's main analysis length is left for the audit's compaction pass (this fire's record replaced sentences and added none).
  • Dedup warnings (informational): the vCenter and PHP entries share entities with earlier entries by product; both are distinct findings (a new flaw and a new release) and declare the nearest predecessors in references[].
  • Reverted update: the record on 2026-10-21/php-critical-release-pre-announced-cve-2026-109618-109619 (an improvement that moved its update target to 8.5.13, 8.4.28, 8.3.37 and 8.2.36 and fixed its CVE-2026-115560 statements) was reverted with git checkout after verifier iteration 5: touching the 81 KB critical entry put the whole file in verification scope, and the passes kept finding vendor lines (Cisco, Fortinet, GreyNoise start date, two evidence quotes, a Siemens version) that its cited pages had changed since it was last written, defects that predate this fire. Left for the quality audit: that entry still says CVE-2026-115560 is unfixed and names 8.5.12, 8.4.27, 8.3.36 and 8.2.35 as the targets, while the new PHP entry states the fix; it also needs the VulnCheck note that its exploit covers PHP 7.4.10 to 7.4.33, 8.0.15 to 8.0.30 and 8.1.10 to 8.1.34.
  • Verifier iteration 5 findings declined: the ENISA record for CVE-2026-132825 re-read at the main agent lists three references (Hacker News, the Ruby announcement, the GHSA) and no GitHub reproduction repository, so the Ruby entry states no public proof of concept; Tenable's updated distribution status for the Java4Shell entry (Ubuntu, Red Hat, Debian fixes) belongs to the previous fire's dated section and is left to the audit.
  • Verification: six iterations, none CLEAN; the loop ended on the low-residual early exit after iteration 6 (truth 2, editorial 0, no F1 or F4): both findings were fixed after that pass (the Java4Shell cves[].fixed Cisco qualifier restored, the PHP "left open in October" clause cited to watchTowr) and have not had a further pass, so verification_residual_count is 2. Iteration 1 dropped the F5 BIG-IP entry; iteration 5 led to reverting the October PHP entry's update. The later passes read pages that had changed since the text was written (Zammad 7.2.3, GitLab's Ruby fix, Cisco's updated product table), which is why the loop did not converge on CLEAN.

← Operations dashboard · run-record contract: docs/pipeline.md