2026-11-05T0255Z-intel
One pipeline fire, in full · intel run of 2026-11-05 · 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/2026-11-05/2026-11-05T0255Z-intel.md.
Run telemetry
- Items returned
- 16
- Duration
- 32m 25s
- Tool calls
- 6 WebFetch12 WebSearch170 bridge
- Cited sources
- 6 of 25 in slice
- Items returned
- 7
- Duration
- 21m 01s
- Tool calls
- 1 WebFetch30 WebSearch95 bridge
- Cited sources
- 4 of 21 in slice
- Items returned
- 10
- Duration
- 32m 11s
- Tool calls
- 4 WebFetch18 WebSearch115 bridge
- Cited sources
- 3 of 17 in slice
- Items returned
- 7
- Duration
- 21m 16s
- Tool calls
- 0 WebFetch31 WebSearch95 bridge
- Cited sources
- 1 of 12 in slice
- Items returned
- 3
- Duration
- 21m 45s
- Tool calls
- 0 WebFetch7 WebSearch53 bridge
- Cited sources
- none
- Items returned
- 3
- Duration
- 14m 18s
- Tool calls
- 0 WebFetch10 WebSearch30 bridge
- Cited sources
- none
- Items returned
- 3
- Duration
- 21m 27s
- Tool calls
- 0 WebFetch2 WebSearch70 bridge
- Cited sources
- none
- Items returned
- 3
- Duration
- 13m 28s
- Tool calls
- 0 WebFetch1 WebSearch17 bridge
- Cited sources
- none
Verification
Deep dive
·
Entries this run published (6) and updated (5)
- CVE-2026-7273 and CVE-2026-107441, Zyxel GS1900 switches: GreyNoise caught a Red Heron-linked actor exfiltrating configurations and hashed root credentials from 996 switches in 48 countries, and Binarly shows its exploit targets a separate flaw that the June fix leaves open; only the .3C0 builds close it
- Danfoss AK-SM and AK-SC, Copeland xweb and Dixell xweb refrigeration controllers compromised, with forced defrost at hundreds of French supermarkets, refrigeration failures at more than 115 shops in at least six more countries and first cases in Germany, and on 2026-11-02 altered settings in the cold rooms of the French blood service through a refrigeration supplier; responsibility not established; ENISA and Europol call for immediate action
- CVE-2026-111497, Node.js and Google Brotli (Node2Shell): an out-of-bounds write in the Brotli decoder lets crafted compressed data run code, reached through fetch() and Express 5 or body-parser 2 request bodies; fixed in Node.js 22.23.4, 24.22.1 and 26.12.1 and in the PyPI packages Brotli 1.2.1 and brotlicffi 1.2.1.0 (CVSS 4.0 9.2)
- CVE-2026-98432, Linux kernel IPv4 TCP: a use-after-free in connection setup, present since 4.4, has an Anthropic-built remote root exploit against MPTCP listeners that CloudLinux could not reproduce; Citrix XenServer, XCP-ng, MikroTik, Red Hat, Debian, CIQ and Amazon Linux have fixes, Ubuntu's are in progress and WatchGuard and Fortinet have none for their products (CVSS 9.8)
- CVE-2026-109618 / CVE-2026-109619, Xerox AltaLink, EC81xx, VersaLink and WorkCentre printers and Fiery print controllers: Xerox lists them as affected by the PHP2Shell flaws, with fixes expected on 2026-11-06 and Fiery patches between 2026-11-09 and 2026-11-23
- Canton Vaud warns of a device-code phishing campaign against Microsoft 365 accounts running in French-speaking Switzerland, with several organisations already affected
- SpecterOps: a Windows iSCSI Target Server set up without CHAP lets any host that knows the target and an authorised initiator name read the shared disks below file permissions, a Hyper-V domain controller's among them
- Sophos's own root-cause analysis of a developer-targeted intrusion: a fake job-interview repository's VS Code workspace settings ran BeaverTail and InvisibleFerret on a Mac whose endpoint agent had not reported since 11 June, and a third-party OAuth authorization outlived the session revocation
- CVE-2026-114424 and CVE-2026-114425, SEPPmail Secure Email Gateway before 15.1.1: an unauthenticated mail sender can reach root command execution through a replacement rule and root file access through Large File Transfer, with three more flaws fixed in the same release and no fix for the 15.0 branch (CVSS 4.0 9.2)
- Sekoia: an APT29-linked watering hole serves n-day iOS and Android exploit chains through compromised sites, clone domains and a one-week compromise of the Plerdy tracker; CISA lists the Apple WebKit pointer-authentication bypass CVE-2024-27834 as exploited on 2026-11-04
- CVE-2026-109593 and CVE-2026-109594, SonicWall GMS 9.5.2 and earlier: an unauthenticated network attacker may read internal GMS files (CVSS 3.1 9.1), and SonicWall says the end-of-life product gets no fix and has no workaround
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.
17 last_successful_fetch · 12 notes · 3 added · 1 status · 1 content_scope.
| Source | Change | From → To | Reason |
|---|---|---|---|
| enisa-euvd | last_successful_fetch | · → 2026-11-05 | fetched and used |
| cisa-kev | last_successful_fetch | · → 2026-11-05 | fetched and used |
| watchguard-psirt | last_successful_fetch | · → 2026-11-05 | fetched and used |
| mikrotik-routeros-changelog | last_successful_fetch | · → 2026-11-05 | fetched and used |
| splunk-advisories | last_successful_fetch | · → 2026-11-05 | fetched and used |
| enisa | last_successful_fetch | · → 2026-11-05 | fetched and used |
| csirt-acn-it | last_successful_fetch | · → 2026-11-05 | fetched and used |
| sekoia | last_successful_fetch | · → 2026-11-05 | fetched and used |
| sonicwall-psirt | last_successful_fetch | · → 2026-11-05 | fetched and used |
| specterops | last_successful_fetch | · → 2026-11-05 | fetched and used |
| sophos-xops | last_successful_fetch | · → 2026-11-05 | fetched and used |
| citrix-netscaler-security-bulletins | last_successful_fetch | · → 2026-11-05 | fetched and used |
| heise-sec | last_successful_fetch | · → 2026-11-05 | fetched and used |
| srf-news | last_successful_fetch | · → 2026-11-05 | fetched and used |
| bleepingcomputer | last_successful_fetch | · → 2026-11-05 | fetched and used |
| xerox-security-bulletins | last_successful_fetch | · → 2026-11-05 | fetched and used |
| vd-ch-press-releases | last_successful_fetch | · → 2026-11-05 | fetched and used |
| ox-security | notes | · → recipe note appended | 2026-11-05 (S3): the page URLs stay 403 on every transport; `python3 tools/fetch_source.py url https://www. |
| sekoia | notes | · → recipe note appended | 2026-11-05 (S3): `url https://www.sekoia.com/blog` gives dated cards and `extract` reads posts, but `extrac |
| cisa-advisories | notes | · → recipe note appended | 2026-11-05 (S1): WebFetch on cisa.gov returned 403 on the listing, the alert pages and all.xml on this date |
| sonicwall-psirt | notes | · → recipe note appended | 2026-11-05 (S1/FU1): the default reader body and `extract` omit the Affected/Fixed/Credits tables; `jina <URL> |
| xerox-security-bulletins | notes | · → recipe note appended | 2026-11-05 (FU4): the bulletin PDF href exists only in the raw index HTML: `url --direct` on the index, then |
| fuitesinfos | notes | · → recipe note appended | 2026-11-05 (S4): `url https://fuitesinfos.fr/articles` now prints markdown-link hrefs, so the older href re |
| netzwoche | notes | · → recipe note appended | 2026-11-05 (S4): `feed https://www.netzwoche.ch/rss.xml` works |
| cyberinsider | notes | · → recipe note appended | 2026-11-05 (S4): `feed https://cyberinsider.com/feed/` gives dated items |
| bitdefender-threat-debrief | notes | · → recipe note appended | 2026-11-05 (S4): `feed https://www.bitdefender.com/nuxt/api/en-us/rss/businessinsights/` works |
| ransomware-live | notes | · → recipe note appended | 2026-11-05 (S4): API field names are group_name, post_title, discovered |
| trellix | notes | · → recipe note appended | 2026-11-05 (S3): reports sit at /advanced-research-center/threat-reports/<month>-<year>/ and `extract` read |
| seppmail-gateway-revision-history | added | · → status: candidate | Added 2026-11-05 (S1): first-party, dated per-CVE security section for the Swiss mail-encryption gateway (NCSC |
| enisa-cnw | added | · → status: candidate | Added 2026-11-05 (S2): ENISA-kept per-incident index of the advisories every EU CSIRTs Network member issued ( |
| binarly-blog | added | · → status: candidate | Added 2026-11-05 (S1): primary for the Zyxel GS1900 exploit-recovery and patch-bypass finding (blog and adviso |
| cert-eu-blog | status | candidate → active | promotion_due: cited by published entries from 10 distinct runs |
| joomla-security-centre | notes | · → recipe note appended | health sweep unreachable, direct reads succeeded minutes later; transient, no demotion |
| tagesschau | content_scope | · → general-news | source_health flagged the recipe as irrelevant (no security vocabulary): a general news outlet; content_scope set to general-news |
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 tried | Method chain | Status / class | What the agent did instead |
|---|---|---|---|---|
| connectwise-trust-center | https://www.connectwise.com/company/trust/advisories | extract → url → jina | 403 transport-403 direct 403, trafilatura no readable body and the reader relayed the block | status.connectwise.com read with extract: no new security item in the window |
| cisa-advisories | https://www.cisa.gov/news-events/cybersecurity-advisories | webfetch → bridge:feed all.xml → webfetch alert URL → websearch → bridge:cisa csaf-recent | 403 transport-403 WebFetch returned 403 on the listing, the alert pages and all.xml; the reader relayed no items | cisa csaf-recent and cisa-kev worked; WebSearch found no AA-series advisory in the window |
| ico-uk | https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/ | bridge:extract → bridge:url → webfetch → websearch | 403 transport-403 403 on extract, url, the reader and WebFetch | WebSearch found no ICO decision dated in the window; 403 never demotes |
Bridge invocations (this run)
4 bridge calls this run · these are successful bridge fetches (separate from "Coverage gaps" above).
- tools/kev_window_diff.py ×1
- tools/fetch_source.py cisa-kev ×1
- tools/fetch_source.py ncsc-csh recent / post ×1
- 4 scoped cti-research deep reads (FU1 to FU4) with tools/fetch_source.py extract, url --direct, pdf and the jina reader where extract was walled ×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 · 20 findings (truth=7, editorial=8, advisory=5) · Claude Sonnet 5.5 · 25m 03s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | The cited Node.js advisory names only the release lines 26.x, 24.x and 22.x and says EOL versions are always affected; it carries none of the fixed version numbers (grep for 22.23.4, 24.22.1, 26.12.1 | · | |
| F3 claim-not-supported | · | The Ubuntu tracker lists only Ubuntu releases ('Vulnerable work in progress'); it never mentions Debian, CIQ or Amazon Linux (0 matches). HEAD cited the Debian Security Tracker, CIQ and AWS for those | · | |
| F4 hallucinated-fact | · | Not a substring of https://aws.amazon.com/security/security-bulletins/2026-160-aws/ as fetched this iteration (Last Updated 11/04/2026 15:25 PST). The page now says only that 'TLS Listeners on Network | · | |
| F4 hallucinated-fact | · | GreyNoise shows the collector's last act as 'cp /tmp/info /home/web/tmp/info.txt' (staging for retrieval) and says nothing about whether anything persists; the entry's own body says the harvested data | · | |
| F4 hallucinated-fact | · | Sekoia's post names an on-device browser artefact (the iOS chains set a replay-guard cookie VDGH_TOKEN_REF with a one-minute expiry) and the Chrome/Wi-Fi files the Android payload reads; and its IOC t | · | |
| F3 claim-not-supported | · | Citrix change log: 'Initial Publication 2026-10-30', revised 2026-11-03 (remediation for LTSR/SCG); Splunk: Published 2026-10-28, Last Updated 2026-11-04. The entry labels both with the revision date, | · | |
| F3 claim-not-supported | · | The administration-guide Home page carries no 2026-11-05 dateline (page metadata date is 2023-01-01); 2026-11-05 is the pipeline's retrieval date. The fact cited (Firmware version in the System data t | · | |
| F9 surface-contradiction | · | Vates VSA-2026-056 says 'As of 2026-11-03, the kernel package for XCP-ng 8.3 have been updated to address this issue' while the XCP-ng post dated 2026-11-04 says the fix 'is now available'. The releas | · | |
| F8 needs-more-research | · | To keep the analysis from growing, the rewrite removed correct, cited, actionable body content: the version-reading caveat (how to read the version; ELLIO says check the release version) now survives | · | |
| F8 needs-more-research | · | The rewrite dropped the 'how to tell' that sources clearly support: the Go team page says 'As of Go 1.24, Go programs enable MPTCP by default for listen sockets on Linux' (so any current Go service is | · | |
| F5 missing-citation | · | True on the vendor pages I read (Apple 120905, 120896, 120903, 120901, 120902; mikrotik 7.26beta5 CHANGELOG and channel list), but none of those pages is in sources[] or cited inline; only Apple 12089 | · | |
| F16 org-triage | · | A cantonal authority reports an active M365 device-code phishing campaign in French-speaking Switzerland with 'several organisations already affected'; the constituency includes cantonal and communal | · | |
| F17 classification | · | After this update the entry's central new claim (the exploited flaw is CVE-2026-107441, and the .2C0 builds stay exposed to the exploit) rests on Binarly alone; the summary itself says 'Binarly alone | · | |
| F18 action-item-discipline | · | One ~170-word action bundles five separate tasks (inventory, take off internet, replace EOL units, written contractor query, independent temperature checks) and largely restates the body's CERT-FR and | · | |
| F12 single-source-flag-missing | · | The entry's frontmatter carries verification: single-source, not single-source-national-cert. The primary is the vendor's own revision history, so single-source is defensible, but the run-record note | · | |
| F11 editorial-advisory | · | Reader-facing sourcing_note carries fetch-transport vocabulary ('readers', 'text extraction') and runs past two sentences of provenance; keep 'the analysis page is undated; the timeline dates are the | · | |
| F11 editorial-advisory | · | Bulletin v5 is dated 2026-11-03, after the entry was published on 2026-11-02 (it cited v4), and adds new vendor mitigations plus a new action. That is a genuine post-publication development, which a t | · | |
| F11 editorial-advisory | · | Zyxel headline states Binarly's attribution as fact while the summary says 'Binarly alone calls it the exploited zero-day' and BleepingComputer says intent is unclear; iSCSI headline drops the precond | · | |
| F11 editorial-advisory | · | The run registered these 15 product keys (entities_added) but no entry's entities[] links any of them (0 entries reference each); the SonicWall GMS entry carries entities: [] and the iSCSI entry links | · | |
| F11 editorial-advisory | · | The cited Sophos RCA and blog describe a job-lure repository run on one developer Mac; neither describes a supply-chain compromise, so the supply-chain tag is unsupported. | · |
Iteration #2 NEEDS_FIXES · 13 findings (truth=6, editorial=5, advisory=2) · Claude Sonnet 5.5 · 28m 50s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | (low confidence) The cited AWS bulletin (Last Updated Date 11/04/2026 15:25 PM PST) no longer carries the NLB TCP-listener sentence and now says 'Global Accelerator does not support IP options on IPv4 | · | |
| F14 quantifier-without-source | · | (low confidence) The Go team page says 'As of Go 1.24, Go programs enable MPTCP by default for listen sockets on Linux. This default may be changed by setting the "multipathtcp" GODEBUG' and that prog | · | |
| F3 claim-not-supported | · | (low confidence) 'Both' covers Vates and XCP-ng but only the XCP-ng post is cited. Vates VSA-2026-056 says 'A public exploit has been published by Anthropic. It leverages an MPTCP listener ... The und | · | |
| F3 claim-not-supported | · | (low confidence) The cited Zyxel advisory carries only 'Revision history 2026-11-3: Initial release'; the 2026-10-22 fix date is on Binarly's blog ('On Oct. 22, 2026 ... they released new firmware ver | · | |
| F3 claim-not-supported | · | (low confidence) Binarly's blog and advisory never state 'Red Heron' in prose; the blog says its team found the exploit in an open directory on the JITTERLY/SIXZUT C&C infrastructure and links Acronis | · | |
| F4 hallucinated-fact | · | The Xerox record was retyped `update` after iteration 1 (entries/2026-11-02/xerox-fiery-php2shell-printers-fixes-due-6-to-23-november.md: `type: update`, `updated_at: 2026-11-05T03:51:20Z`, section `# | · | |
| F9 surface-contradiction | · | (low confidence) BleepingComputer, cited in the same section, writes 'Coolworld says the attackers exploited a vulnerability in a third-party component'; Coolworld's own statement says only 'The vulne | · | |
| F8 needs-more-research | · | (low confidence) The Akamai page (read in this iteration from the raw HTML) gives the direct listener check the entry still lacks: 'To determine whether any applications are listening on MPTCP sockets | · | |
| F18 action-item-discipline | · | (low confidence) The trim removed only the unsourced 'in writing' clause. The action is still one ~150-word sentence that bundles inventory, taking the interface off the internet, VPN restore, the sto | · | |
| F17 classification | · | (low confidence) Both cited sources are on sophos.com: the blog is the sophos-xops source record, tier B in sources/sources.json, and the RCA page has no source record at all (its page metadata date i | · | |
| F5 missing-citation | · | (low confidence) The cited MikroTik changelogs (upgrade.mikrotik.com/routeros/7.25.1/CHANGELOG and 7.24.6/CHANGELOG) carry only 'system - improve stability (includes CVE-2026-98432)' and no channel; t | · | |
| F11 editorial-advisory | · | Two wording points. (1) 'describes ... and no implant' is literally true (the GreyNoise Zyxel section names none) but reads as GreyNoise asserting absence; 'names no implant' is safer. (2) The takeawa | · | |
| F11 editorial-advisory | · | The Xerox bulletin (v5) makes no statement on exploitation either way; it says the Product Security team is 'reviewing ... whether these vulnerabilities can be exploited in those products'. HEAD's 'gi | · |
Iteration #3 NEEDS_FIXES · 4 findings (truth=2, editorial=1, advisory=1) · Claude Sonnet 5.5 · 22m 04s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | The cited Binarly advisory page carries the 9.8 / AV:N scoring and an 'Exploitation in the wild' section but has no proof-of-concept statement (0 matches for 'proof-of-concept', 'PoC', 'minimi') and d | · | |
| F3 claim-not-supported | · | (low confidence) The Zyxel advisory carries the .2C0-and-earlier affected table and the .3C0 patch column (supports 'still carries the flaw'), but nothing about exploitation or what Binarly says: its | · | |
| F5 missing-citation | · | Iteration 2 finding #11 (channel labels from a NEWEST7 listing that is in no sources[] record) was remediated in cves[].fixed and in the main analysis, but the label survives in the Update section: th | · | |
| F11 editorial-advisory | · | The bulletin (BL02/261104/CSIRT-ITA) gives two priorities, not one: 'dando priorità alla rimozione dell’esposizione diretta su Internet delle interfacce amministrative e di controllo e alla verifica d | · |
Iteration #4 NEEDS_FIXES · 6 findings (truth=5, editorial=1, advisory=0) · Claude Sonnet 5.5 · 14m 39s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F4 hallucinated-fact | · | The cited Red Hat page carries a fix table (client-rendered, in the page payload and the hydra API https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-98432.json): RHEL 8 RHSA-2026:93414 (k | · | |
| F8 needs-more-research | · | Red Hat shipped kernel fixes for RHEL 8, 9 and 10 on 2026-11-01 (RHSA-2026:93414, 93412, 93413; EUS/E4S/SAP streams on 2026-11-02 to 11-04), and the Red Hat page and hydra API show them. The entry nam | · | |
| F13 analytical-link-as-fact | · | (low confidence) No cited source states that CISA acted on GreyNoise's research. The CISA alert lists only 'CVE-2026-7273 Zyxel GS1900 Series Switches Stack-Based Buffer Overflow Vulnerability'; Bleep | · | |
| F3 claim-not-supported | · | The clause attributes a statement to Binarly but cites only Zyxel and BleepingComputer. BleepingComputer carries its half ('an unauthenticated attacker with network access'); the Binarly half ('an att | · | |
| F14 quantifier-without-source | · | (low confidence) BleepingComputer, a cited source, calls it a zero-day in its own voice: title 'Zyxel fixes zero-day exploited to hack nearly 1,000 network switches' and lede 'Zyxel has patched a zero | · | |
| F4 hallucinated-fact | · | (low confidence) The Debian tracker lists 'trixie (security) | 6.12.115-2 | fixed' and carries no date for any row; 2026-11-03 is the date the tracker was read, but the parenthesis reads as the releas | · |
Iteration #5 NEEDS_FIXES · 6 findings (truth=3, editorial=0, advisory=3) · Claude Sonnet 5.5 · 30m 21s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | The citation terminating the clause is Sekoia only; the Sekoia page (cached extract, raw HTML and flattened tables) contains no CISA/KEV text (0 matches for CISA, KEV, Known Exploited). 'CISA lists ev | · | |
| F3 claim-not-supported | · | (low confidence) GreyNoise (https://www.greynoise.io/blog/open-season-on-kapibala-attacker-steals-government-records-wordpress-exploitation) says 'GreyNoise has been tracking malicious use of an IP ad | · | |
| F3 claim-not-supported | · | (low confidence) The Binarly blog and advisory and BleepingComputer say CISA added CVE-2026-7273 to KEV, but none of the three says the catalog does not list CVE-2026-107441. The statement is true: KE | · | |
| F11 editorial-advisory | · | (low confidence, advisory) Both record summaries narrate record-keeping ('the entry's earlier ... replaced') instead of stating the delta. The Zyxel record summary also states that BleepingComputer re | · | |
| F11 editorial-advisory | · | (low confidence, advisory) Reader-facing sourcing_note carries fetch-process narration ('text extraction misses') and runs to four sentences; the provenance can be one clause ('RHEL fix builds from Re | · | |
| F11 editorial-advisory | · | (low confidence, advisory) The 'earlier bulletin' (BL01/261013) is labelled 2026-11-04, the page's last-updated date. The page's version table: 1.0 published 13-10-2026, 1.1 03/11/2026 and 1.2 04/11/2 | · |
Iteration #6 CLEAN · 4 findings (truth=0, editorial=0, advisory=4) · Claude Sonnet 5.5 · 22m 54s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F11 editorial-advisory | · | (low confidence) The Vaud notice prints the structure "[nom]-portail.cloud-storage-id0XXX.workers.dev"; "[nom]" is simply "name". "company-style" is the entry's gloss and the notice does not say the n | · | |
| F11 editorial-advisory | · | (low confidence) The page's own date is 2026-10-30 and its TL;DR now reads "last updated November 3, 2026, 11:44 UTC"; the label 2026-11-02 matches neither. The sources[] record names it as the update | · | |
| F11 editorial-advisory | · | (low confidence) The prior-iteration fix removed the fetch-method narration but left a capital T after the semicolon and the clause "as its security-data API serves it", which is still a trace of how | · | |
| F11 editorial-advisory | · | (low confidence) This run added Splunk and Citrix to an action that already bundled Envoy and VMware, so one list item now carries four vendors' tasks for four different owners. Every clause is concre | · |
Iteration #7 NEEDS_FIXES · 8 findings (truth=1, editorial=0, advisory=7) · Claude Sonnet 5.5 · 27m 09s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F4 hallucinated-fact | · | Three of the four 'after' word counts are stale. The gate's own tools/check_run.py _main_words() on the files as they stand gives Danfoss 736 -> 735, Node2Shell 549 -> 539 (matches), CVE-2026-98432 54 | · | |
| F11 editorial-advisory | · | (low confidence) The notice prints [nom]-portail.cloud-storage-id0XXX.workers.dev; [nom] is 'name', and the notice does not say the name imitates a company. 'company-style' is the entry's gloss; 'a na | · | |
| F11 editorial-advisory | · | (low confidence) The body cites Pen Test Partners only on the Exposure clause (port 3260); the segmentation advice in the Defender takeaway is cited to SpecterOps alone. The PTP post does carry segmen | · | |
| F11 editorial-advisory | · | (low confidence) 'no fetched Apple page' narrates how the research was done inside a reader-facing field. Say 'no Apple security-content page read lists a fix for iOS 16.7.x' or state it as the open q | · | |
| F11 editorial-advisory | · | (low confidence) Stray capital after the semicolon and a residual trace of how the page was read ('as its security-data API serves it'). One clause of provenance ('the fixed RHEL builds are those of t | · | |
| F11 editorial-advisory | · | (low confidence) The AWS bulletin shows Publication Date 10/30/2026 and Last Updated 11/04/2026 (the version read on 2026-11-03 showed Last Updated 11/02/2026), so 'as published on 2026-11-02' should | · | |
| F11 editorial-advisory | · | (low confidence) All five flaws are unauthenticated on the CNA records (EUVD-2026-116781, -116782, -116796, -116798, -116797 each say unauthenticated remote sender or attacker), and the entry says so | · | |
| F11 editorial-advisory | · | (low confidence) One list item carries four vendors' tasks for four different owners. Every clause is concrete and source-backed (Envoy GHSA, Broadcom VMSA-2026-0008.1, Splunk SVD-2026-1015, Citrix CT | · |
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
2026-11-05T0255Z-intel · Sonnet 5.5 · window 26 h · 6 entries published
Verification & coverage notes
- Coverage window: standard, 26 h (previous intel run 2026-11-04T0256Z-intel, gap 24.0 h); no outage backfill. Source rotation ranked on the cursor; the previous two fires' attempts excluded 27 (S1), 31 (S2) and 37 (S3) standard records, and the exclusion was waived for S4 because all 14 breach-domain sources had been attempted by those fires.
- The entity registry (704 KB) was grepped by key and alias, not read whole; the 14-day coverage index, the state digest and the taxonomy were read in full. S5 intake: no closed-source drops.
- Backlog: the ARX Group SA row stays held (S4 re-checked the ransomware.live record, a breach tracker, the ARX site, German and Italian searches and the Inside IT, Netzwoche and SRF feeds: no confirmation; expiry 2026-11-07 unchanged). The ChimeraZ row stays held (no commune, canton, zone or Swiss press confirmation; FrenchBreaches still calls the hack unconfirmed; expiry 2026-11-17 unchanged). Neither row changed state, so nothing is appended to them. The Amadeus hold of 2026-11-04 stays dropped (six Asian carriers, no European confirmation).
- KEV sweep (work/2026-11-05T0255Z-intel/kev-window.txt): one addition in the window, CVE-2024-27834, NOT COVERED, published as the new entry on the Sekoia APT29-linked watering hole. No RANSOMWARE row for a covered CVE appeared in the window. S1's diff against the 2026-07-29 snapshot found 26 older flips; the one that touches covered ground, CVE-2025-8088 (WinRAR, ClamAV's bundled UnRAR), is a borderline-drop below.
- borderline-drop: S2-2 Federal Supreme Court blocking ruling: no obligation change for defenders
- borderline-drop: S2-3 EFK audit of BACS reporting platform: oversight, no change to reporting obligation
- borderline-drop: S2-4 Eye-Able supplier incident: no public-sector customer shown, no vector
- borderline-drop: S2-5/S4-2 Holcim + Emperador: private multinational, no vector or TTP (incident floor)
- borderline-drop: S2-6 SiK-N motion 26.4717: filing-stage motion, no obligation
- borderline-drop: S3-1 BSI/Apple threat notifications: out-of-window (BSI 2026-11-02, heise 2026-11-03), no Swiss recipient reported
- borderline-drop: S3-2 Trellix CyberThreat Report: vendor telemetry and leak-site counts, no new defender decision
- borderline-drop: S3-5/S4-3 AI platform credential-theft wave (WorkOS, Chainstack, Wiz, CSA, Mandiant LLMjacking): no constituency victim, platform-by-platform root causes
- borderline-drop: S3-6 Elastic threats etc report: landing page only readable, claims not verifiable
- borderline-drop: S3-7 Report URI Polymarket/DebugBear: crypto platform, out-of-nexus, generic SRI/CSP lesson (lookback, dated 2026-10-30)
- borderline-drop: S3-8 DHS Acronis order: US measure, no Swiss or EU obligation, dated 2026-10-30
- borderline-drop: S3-9 Calif.io Lockdown Mode bypass: video only, no technical detail or second source
- borderline-drop: S1-4 Xen XSA-523: guest-to-host needs prior guest compromise, no exploitation, normal patch cycle
- borderline-drop: S1-5 APT CVE-2026-109518: needs on-path attacker or malicious mirror, no exploitation, normal patch cycle
- borderline-drop: S1-6 HPE Fabric Composer HPESBNW05172: internal-management niche product, fixed, no exploitation
- borderline-drop: S1-7 WatchGuard AuthPoint Gateway: availability and replay impacts, no RCE, no exploitation
- borderline-drop: S1-12 Rails Active Storage CVE-2026-122006: three exposure conditions, no exploitation
- borderline-drop: S1-13 NoMachine CVE-2026-114549: ZDI disclosure of 2026-10-22, fixed since July/October, no exploitation
- borderline-drop: S1-14 Quest Foglight for Databases: ZDI sole source, no vendor advisory read, niche
- borderline-drop: S1-15 Android November bulletin: published 2026-11-02, no exploitation noted, monthly cycle
- borderline-drop: S4-1 Solnet webmail logs: single outlet, unnamed flaw, no public-sector customer shown (incident floor)
- borderline-drop: S4-4 Japan MOFA: primary dated 2026-10-30, outside window
- borderline-drop: S4-5 Credit Mutuel Arkea: financial sector, single outlet, no constituency nexus
- borderline-drop: S4-6 Iceland Inna school system: single secondary source, early-stage, no Swiss link
- borderline-drop: Amadeus booking-display compromise: Asian carriers only, no European confirmation
- borderline-drop: Chrome 157/Edge 213 fixes none exploited; Dameware, GLPI, ManageEngine, Dell, StorageGRID, WSO2 etc: normal patch cycle (S1 drops)
- borderline-drop: CVE-2025-8088 ransomware flag, the KEV record reads Known in catalog 2026.08.07 and Unknown in 2026.07.29 (FU4 bracketed the flip from committed snapshots), so it is not a development of this window; CISA names no group, and no source names a ransomware actor using the flaw. Four store entries cite the CVE and none mentions ransomware; the 2026-10-29 MDaemon and ClamAV entry is the natural home if the audit judges it worth a line. FU4 also found two older defects for the audit: the 2026-06-02 Sekoia/Gamaredon entry dates the WinRAR 7.13 fix to "August 2025" while ESET says 2025-07-30, and the 2026-06-27 Turla STOCKSTAY entry says "Patch WinRAR to 7.11+" while ESET names 7.13 as the fix.
- Single-source and carve-outs: the Vaud device-code notice (single-source, first-party cantonal authority, no corroboration found in nine German and French searches and the NCSC.ch hub); SEPPmail (single-source: NCSC.ch is the CNA and the vendor page describes the same flaws, one assessor); the APT29-linked watering hole (single-source Sekoia; The Hacker News and BleepingComputer relay it); the Windows iSCSI Target Server research (single-source SpecterOps, Pen Test Partners 2016 only for the port); Sophos RCA INC-2026-002 (single-source-victim); SonicWall GMS (single-source vendor, CERT-FR relays it).
- Contradictions disclosed in the entries and not resolved: Zyxel calls the attacker LAN-based while Binarly and BleepingComputer say network access suffices, and Binarly, with BleepingComputer repeating it, calls CVE-2026-107441 an exploited zero-day on the strength of the exploit match while Zyxel and CISA do not connect the activity to it; Vates, XCP-ng and WatchGuard say a public exploit exists or circulates while Akamai knows of none and the MPTCP exploit does not run on XCP-ng or Fireware OS; SEPPmail's vendor page lists the CVSS scores of CVE-2026-114420 and CVE-2026-114421 the wrong way round against the NCSC.ch CNA records.
- Advisory items left for the next quality audit (low-confidence F11, each marked may-be-left by the verifier): the Vaud wording of the notice's company placeholder, the CloudLinux and AWS date labels on the kernel entry, the iSCSI sourcing note's Pen Test Partners remark, the APT29 CVE-2025-43529 fixed-field wording, the kernel sourcing-note capital, the SEPPmail 'two are unauthenticated' phrasing and the four vendor tasks bundled in Node2Shell actions[1].
- Reduced confidence: none. Entries whose primary needed a non-default transport: the Binarly pages (rendered from the page payload), the SonicWall PSIRT tables (reader, html mode) and the Sophos RCA analysis sections (raw page; extract truncates them), so a verifier reading them with extract will not see those passages.
- Deep read: the will-publish set was 12 items (6 new, 6 changelog records), above the eight the prompt names, so the deep read ran as four scoped cti-research follow-ups (FU1 to FU4) after Phase 1 returned; the main agent composed from their claim tables. One follow-up (FU2) reports a classifier stop on its first write of the SpecterOps attack steps and rebuilt the item at awareness level, which is the level the entry keeps.
- One planned changelog record was dropped after the deep read (CVE-2025-8088, above), leaving 5 updates: Danfoss (update), Zyxel GS1900 (update), CVE-2026-98432 (update), Node2Shell/Splunk (update) and the Xerox PHP2Shell entry (update: bulletin v5 of 2026-11-03 is a post-publication development, so it floats).
- Entry rewrites replaced text instead of adding it: the Danfoss (736 to 735 words), Node2Shell (549 to 539), CVE-2026-98432 (546 to 546) and Zyxel (641 to 613) analyses did not grow.
- Sources: three new candidates (seppmail-gateway-revision-history, enisa-cnw, binarly-blog), each with its reason in sources_changed[]; cert-eu-blog promoted to active on the digest's promotion_due (10 contributing runs); recipe notes appended for ox-security, sekoia, cisa-advisories, sonicwall-psirt, xerox-security-bulletins, fuitesinfos, netzwoche, cyberinsider, bitdefender-threat-debrief, ransomware-live and trellix. The CISA recipe note records that WebFetch returned 403 on every cisa.gov HTML path today, so the 2026-10-21 and 2026-10-23 notes that WebFetch reads the listing no longer hold.
- Coverage gaps: ico-uk (403 on every transport, WebSearch found nothing in the window); connectwise-trust-center (403, status page substituted); cisa-advisories HTML pages (403; cisa-kev and cisa csaf-recent worked and WebSearch found no AA-series advisory); ibm-support-security-bulletins (not attempted, no listing transport; a relay sweep through BSI found only a QRadar XSS and a WebSphere DoS); Inside IT body (paywalled, in the page's embedded data); cyber.gouv.fr RSS (XML parse failure); securityweek and therecord feeds too short for the 168 h lookback; crowdstrike feed carries product news only; Rapid7 feed unparseable.
← Operations dashboard · run-record contract: docs/pipeline.md