2026-10-28T0256Z-intel
One pipeline fire, in full · intel run of 2026-10-28 · 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/2026-10-28/2026-10-28T0256Z-intel.md.
Run telemetry
- Items returned
- 8
- Duration
- 31m 01s
- Tool calls
- 11 WebFetch19 WebSearch120 bridge
- Cited sources
- 6 of 17 in slice
- Items returned
- 6
- Duration
- 28m 10s
- Tool calls
- 2 WebFetch37 WebSearch95 bridge
- Cited sources
- 1 of 22 in slice
- Items returned
- 11
- Duration
- 24m 21s
- Tool calls
- 3 WebFetch18 WebSearch150 bridge
- Cited sources
- 2 of 14 in slice
- Items returned
- 9
- Duration
- 34m 30s
- Tool calls
- 2 WebFetch41 WebSearch95 bridge
- Cited sources
- 0 of 14 in slice
Verification
Deep dive
·
Entries this run published (5) and updated (9)
- CVE-2026-61425, Balbooa Gridbox for Joomla: a client-supplied cookie is accepted as proof of identity, giving anonymous Super User access
- Swiss commune Manno (TI) confirms a cyberattack that encrypted part of its servers on 4 August 2026; SafePay has now published about 30 GB, which Risikomonitor reads as social-assistance decisions with AHV numbers, naturalisation files and a near-complete address book
- IDC Frontier: ransomware stops four zones of IDCF Cloud for 495 companies and local governments; the provider's seventh report traces it to an open-source flaw in the customer portal's authentication system and says no server data left
- CVE-2026-109618 / CVE-2026-109619, PHP (PHP2Shell): two unauthenticated remote-code-execution flaws in the runtime, fixed on 2026-10-21 in 8.5.12, 8.4.27, 8.3.36 and 8.2.35, with attack traffic already observed and no upstream release for 8.1 and older (CVSS 4.0 9.2)
- CVE-2026-109618 / CVE-2026-109619, Palo Alto Networks PAN-OS and Panorama: the PHP2Shell flaws affect every release through the management web interface, and Palo Alto reports active exploitation against internet-exposed management interfaces; fourteen hotfixes are out and six are due by 2026-10-30
- CVE-2026-109618 / CVE-2026-109619, NetScaler ADC and Gateway: the PHP2Shell flaws reach the management GUI on the NSIP, management-enabled SNIPs and the cluster IP of every supported version; fixed in 14.1-73.49, 13.1-64.32 and the FIPS and NDcPP builds
- 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-2026-111497 / CVE-2026-70197, Microsoft .NET: an out-of-band release fixes a Brotli decoder flaw and delivers the Windows sockets fix that the 13 October update did not contain, in 8.0.33, 9.0.22 and 10.0.14 (CVSS 3.1 8.1)
- Camping.Care booking system (TCS Camping): customer and booking data stolen in connection with the PHP flaws, with real reservation numbers already used in WhatsApp fraud and 15,000 guests warned as a precaution
- CVE-2026-109618 / CVE-2026-109619, Cisco Expressway, Ultra-Reliable Wireless Backhaul, Duo Access Gateway and ISE Portal Builder: Cisco's final PHP2Shell advisory names Expressway 15.5.21 and Duo Access Gateway 3.2.0 as the fixes and knows of no malicious use against its products
- CVE-2026-109618 / CVE-2026-109619, HPE Networking ClearPass Policy Manager: bulletin HPESBNW05171 lists 6.14.1 and 6.11.16 and below as affected by the PHP2Shell flaws and names hotfix patches for both (CVSS 3.1 9.8)
- ARA-Region Lyss-Limpachtal (canton Bern, communal wastewater-treatment association): SafePay has put about 49 GB into the darknet, which Risikomonitor's analysis, relayed by Inside IT, finds holds personal records, bank statements, weak passwords and a third party's monitoring credentials; the association declines to comment
- CVE-2026-48356 (PolyShell), Adobe Commerce and Magento Open Source: an unauthenticated file upload attacked since March 2026, fixed for supported lines in July, and listed by CISA as exploited on 2026-10-27
- DeadTag: re-registered domains of defunct analytics and advertising tags redirect visitors of thousands of sites, among them to a page that serves the Coruna and DarkSword iPhone exploit kits, which compromise a vulnerable device when the page opens
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.
9 last_successful_fetch · 7 notes · 2 consecutive_fetch_failures · 2 added · 1 status · 1 health_cmd.
| Source | Change | From → To | Reason |
|---|---|---|---|
| cisa-kev | last_successful_fetch | 2026-10-27 → 2026-10-28 | fetched and used |
| cisa-advisories | last_successful_fetch | 2026-10-27 → 2026-10-28 | fetched and used |
| paloalto-psirt | last_successful_fetch | 2026-10-27 → 2026-10-28 | fetched and used |
| hpe-networking-psirt | last_successful_fetch | 2026-10-09 → 2026-10-28 | fetched and used |
| mysites-guru | last_successful_fetch | 2026-10-23 → 2026-10-28 | fetched and used |
| tenable-research | last_successful_fetch | 2026-09-30 → 2026-10-28 | fetched and used |
| inside-it-ch | last_successful_fetch | 2026-10-23 → 2026-10-28 | fetched and used |
| iverify-blog | last_successful_fetch | · → 2026-10-28 | fetched and used |
| report-uri-blog | last_successful_fetch | 2026-10-23 → 2026-10-28 | fetched and used |
| connectwise-trust-center | consecutive_fetch_failures | 5 → 6 | 403 on direct, extract, url, the reader and WebFetch again; a 403 never demotes |
| ec-digital-strategy-newsroom | consecutive_fetch_failures | 1 → 2 | 403 (CloudFront) on every transport for the third run; a 403 never demotes |
| hpe-networking-psirt | notes | · → recipe note appended | 2026-10-28 (S1): recipe note from the sub-agent source ledger, text in sources.json |
| cisco-psirt | notes | · → recipe note appended | 2026-10-28 (S1): recipe note from the sub-agent source ledger, text in sources.json |
| paloalto-psirt | notes | · → recipe note appended | 2026-10-28 (S1): recipe note from the sub-agent source ledger, text in sources.json |
| inside-it-ch | notes | · → recipe note appended | 2026-10-28 (S2): recipe note from the sub-agent source ledger, text in sources.json |
| apa-ots | notes | · → recipe note appended | 2026-10-28 (S2): recipe note from the sub-agent source ledger, text in sources.json |
| ransomware-live | notes | · → recipe note appended | 2026-10-28 (S4): recipe note from the sub-agent source ledger, text in sources.json |
| databreaches-net | notes | · → recipe note appended | 2026-10-28 (S4): recipe note from the sub-agent source ledger, text in sources.json |
| karimi-honeypot-data | status | candidate → active | sources.promotion_due: cited by published entries from at least three distinct runs |
| ellio-threat-research | added | · → candidate | first-hand exploitation telemetry cited by a published entry |
| jfrog-security-research | added | · → candidate | original research cited by a published entry |
| bfv-verfassungsschutz-de | health_cmd | · → feed recipe of the RSS URL | health probe of the landing page answered HTTP 400 while extract and the feed recipe read cleanly; the check now runs the feed recipe |
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 | bridge:url → bridge:extract → jina → webfetch | 403 transport-403 403 on direct, extract, url, the reader and WebFetch; no readable body (failing on every run since 2026-10-19) | status.connectwise.com read with extract (only the covered 2026-10-16 incident) and a WebSearch for October 2026 advisories found nothing new |
| ec-digital-strategy-newsroom | https://digital-strategy.ec.europa.eu/en/news | bridge:extract → bridge:url → jina → webfetch | 403 transport-403 403 (CloudFront) on extract, the reader fallback, url and WebFetch | Commission policy items probed by WebSearch; nothing dated in the window surfaced |
| resecurity | https://www.resecurity.com/blog | bridge:extract → webfetch → websearch | 200 js-shell extract returns an empty JavaScript shell; no in-container reader for the listing | WebSearch restricted to resecurity.com lists dated article pages; the newest post is 2026-10-14, outside the horizon |
| golem-security | https://www.golem.de/specials/security/ | bridge:extract → bridge:url | 200 cookie-wall listing sits behind a cookie wall for every in-container transport | heise-sec and Inside IT cover the same DACH news |
Bridge invocations (this run)
3 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 extract / url --direct ×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 · 24 findings (truth=8, editorial=10, advisory=6) · Claude Sonnet 5.5 · 29m 58s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | Live Cisco page (https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-php-october26-Jj384Daq, v1.8) has the footnote marker on the URWB row: <td>Ultra-Reliable Wirele | · | |
| F3 claim-not-supported | · | CSAF https://csaf.arubanetworking.hpe.com/2026/hpe_networking_-_hpesbnw05171.json workaround text: "until ClearPass instances have been upgraded to a version providing some level of mitigation. These | · | |
| F1 broken-url | · | WebFetch returns HTTP 404 and the bridge returns only the site navigation (11 KB, no article text); no other source read this iteration states 'the websites of Ibaraki Prefecture and the city of Kodai | · | |
| F14 quantifier-without-source | · | https://www.idcf.jp/news/topics/20261027001 is titled 【第7報】 (seventh report) and says 本件に関する一連の調査が完了しました (the series of investigations is complete); it never calls itself the final report and announce | · | |
| F5 missing-citation | · | No cited source carries a TCS statement tying the theft to PHP: SRF writes the PHP link in its own voice ("Im Zusammenhang mit der weltweiten Sicherheitslücke bei der Programmiersprache PHP warnt TCS | · | |
| F3 claim-not-supported | · | (low confidence) Report URI: "Search engines, security scanners, automated browsers and anything else that looks like a bot gets nothing." and "When we investigated from data-centre IP addresses, or c | · | |
| F3 claim-not-supported | · | (low confidence) The clause ends on the .NET Blog cite; the blog (https://devblogs.microsoft.com/dotnet/dotnet-october-2026-oob-security-updates/) describes the OOB release and says nothing about expl | · | |
| F14 quantifier-without-source | · | (low confidence) The Cisco table names six items (URWB, Unified Industrial Wireless Software, Expressway Series, Expressway Select Series, Duo Access Gateway, ISE Portal Builder), the entry's own affe | · | |
| F4 hallucinated-fact | · | (low confidence) The advisory says the bugs "contain additional platform-specific information, including workarounds (if available) and fixed software releases"; it does not say affected-version range | · | |
| F5 missing-citation | · | The sentence has no citation and reads as Balbooa's account, but Balbooa's page only says "Implemented strict whitelist validation against getBlogPostsSortingList() for the sort-by URL parameter"; the | · | |
| F7 drop | · | (low confidence) Read-side unauthenticated SQLi in a third-party Joomla extension: no exploitation, no public PoC, no Swiss or public-sector exposure shown (regions global, sectors empty). The run's o | · | |
| F8 needs-more-research | · | The advisory lists "Action Links for This Advisory: Snort Rule 67359, Snort Rule 67360". Cisco does publish detection content for the flaws; name the Snort rule ids as a network-detection option. | · | |
| F8 needs-more-research | · | Now that the vendor reports exploitation against internet-exposed management interfaces, the actionability contract wants the post-patch compromise check. The KEV record the entry cites (https://www.c | · | |
| F8 needs-more-research | · | The entry's own 2026-10-27 section says NCSC names Bun among the affected dependents, yet Bun is absent from cves[].affected and affected_products. ELLIO's fix table (https://ellio.tech/en/blog/node2s | · | |
| F8 needs-more-research | · | iVerify (https://www.iverify.com/blog/malvertising-coruna-darksword-iphone-exploits): "Based on current findings, DeadTag appears to be mostly focused on European domains versus .com". That is the onl | · | |
| F8 needs-more-research | · | (low confidence) The seventh report adds: "お客さまのログインパスワードが最大4件、不正に変更されていたことを確認しました。該当するパスワードはすでにリセットしており" (up to four customer login passwords were improperly changed and have been reset). A tampered | · | |
| F9 surface-contradiction | · | The vendor timeline (security.paloaltonetworks.com/json/CVE-2026-109618) ends with 2026-10-27T16:40 'Severity Updated to Critical and Urgency updated as Highest' and carries no later entry. The Waybac | · | |
| F18 action-item-discipline | · | (low confidence) Restates the entry's own Detection and Defender takeaway (inventory external scripts, flag re-registered domains, enforce a CSP) as a task, i.e. check 10b(b); the CSP half is generic | · | |
| F11 editorial-advisory | · | Both rest on one outlet (Inside IT) relaying Risikomonitor's unverified analysis (entries rated B3 and C3); the body hedges, the headline does not. Add 'reports Inside IT' / 'per Risikomonitor' or cut | · | |
| F11 editorial-advisory | · | Detection lines and sourcing_notes point readers to 'the PHP entry' / 'the entries listed under references', and the PHP 2026-10-28 section says 'the PAN-OS entry carries the hotfix table'. Describe t | · | |
| F11 editorial-advisory | · | Eleven paragraphs of main analysis before the labelled lines plus six dated sections, for a non-deep-dive entry whose update this fire is one vendor statement. The vendor roundup paragraph and the per | · | |
| F11 editorial-advisory | · | cves[].status now carries exploited and the headline reports active exploitation, but tags (checked by the filter and brief) lack actively-exploited, which the PHP, Node2Shell and .NET entries carry; | · | |
| F11 editorial-advisory | · | (low confidence) Neither Report URI nor iVerify names CVE-2026-86950 or ties Coruna or DarkSword to it; the link reads as an analytical connection without a source. Drop it or state in the body what c | · | |
| F11 editorial-advisory | · | The seventh report says 206 customers' backups are retrievable (confirmed 2026-10-20). The dated 194 figure is historically right, but the main analysis and Defender takeaway never mention the later c | · |
Iteration #2 NEEDS_FIXES · 20 findings (truth=12, editorial=4, advisory=4) · Claude Sonnet 5.5 · 30m 35s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | The cited CVE JSON (fetched this iteration) has no ETA dates for the nine unreleased builds and no Brotli text (0 matches for 'brotli'; 'h17' occurs only inside CPE strings). The HTML advisory https:/ | · | |
| F3 claim-not-supported | · | The JSON solutions text lists the suggested builds with no estimated dates; the dates (10/27 to 10/30) are only in the HTML table of https://security.paloaltonetworks.com/CVE-2026-109618. The HEAD ver | · | |
| F3 claim-not-supported | · | Both cites were HTML URLs in HEAD and were switched to the JSON URL in this run's diff. The JSON carries no FAQ (no 'brotli' text), and its exploits field now reads 'Active exploitation has been obser | · | |
| F3 claim-not-supported | · | The fifth report is dated 2026-10-14 and carries the 194 figure; a count confirmed on 2026-10-20 cannot be on it. The seventh report (https://www.idcf.jp/news/topics/20261027001) has '10月20日 | 206のお客さ | · | |
| F3 claim-not-supported | · | The cited VCS advisory carries 'X15.5.3 and earlier ... X15.5.20' and CVE-2026-76466 but not 15.5.21, which is the table value on cisco-sa-php-october26-Jj384Daq. The clause chains a fact from each ad | · | |
| F3 claim-not-supported | · | The PHP advisory never mentions X15.5.20 or the clustered-device flaw (that is cisco-sa-vcs-static-cred); the clause is an inference across two advisories under one cite. Add the VCS advisory cite and | · | |
| F3 claim-not-supported | · | The only cite on the sentence is mySites.guru, whose page names no function (0 matches for getBlogPosts or GridboxHelper). The allowlist and the method name are on Balbooa's page: 'strict whitelist va | · | |
| F3 claim-not-supported | · | The sentence's first clause (the PAN advisory lists every release as affected) ends on a PHP citation, and the next sentence's PAN cite states only the management-interface and GlobalProtect facts. Pu | · | |
| F3 claim-not-supported | · | HPE says only 'customers are asked to consider removing guest portals and all management interfaces from public access'; nothing about guest portals being user-reachable by design. Drop the clause or | · | |
| F3 claim-not-supported | · | HPE names 'guest portals and all management interfaces' together and ranks nothing as first. The clause also repeats the preceding clause ('treat any guest portal or management interface ... as the fi | · | |
| F4 hallucinated-fact | · | The advisory table gives only 'Fixed Release Available' (15.5.21, 3.2.0) and no affected range; the same sentence says so. Asserting 'below 15.5.21' and 'below 3.2.0' as the exposed population is the | · | |
| F4 hallucinated-fact | · | The original field stops at '(TCS)'; the English adds 'confirmed', the translation of 'am Montag auf Anfrage der Nachrichtenagentur Keystone-SDA bestätigte', which is not in the original string. Exten | · | |
| F7 drop | · | The ground holds only narrowly. (1) PD-11(b)'s otherwise limb names 'sibling flaws in the same wave weaponised within days'; the July flaws were a different wave three months earlier, and the entry it | · | |
| F8 needs-more-research | · | Tenable's table (https://www.tenable.com/blog/brotli2shell-frequently-asked-questions-for-google-brotli-decoder-cve-2026-111497) gives Istio '< 1.29.9, < 1.30.6, < 1.31.2'; ELLIO's list gives the Debi | · | |
| F10 missed-angle | · | CISA's alert (https://www.cisa.gov/news-events/alerts/2026/10/27/cisa-adds-two-known-exploited-vulnerabilities-catalog) lists it 'based on evidence of active exploitation' and the KEV record shows due | · | |
| F10 missed-angle | · | prompts/cti-run.md PD-11 requires a borderline-drop line for every borderline drop; the run record lists only CVE-2026-48356. Moldova (public-sector, Europe, unauthenticated service endpoints, about 7 | · | |
| F11 editorial-advisory | · | The iteration-1 F11 (reader-facing cross-references to the store's own entries) was fixed for the Cisco entry and the Detection lines but these three sourcing_note sentences still point readers at 'th | · | |
| F11 editorial-advisory | · | Partly fixed in iteration 1 (57 words cut); the per-reporter exploitation chronology and the vendor roundup paragraph still read as a deep dive on a deep_dive: false entry. Candidate for the audit. | · | |
| F11 editorial-advisory | · | The reuse looks benign (same kit, different operators; Report URI: 'Coruna and DarkSword are separate exploit kits that have been seen in different campaigns'), so not an F15, but a defender who read | · | |
| F11 editorial-advisory | · | The entry labels were corrected to 'seventh report' after iteration 1 (the page is titled 【第7報】 and announces the portal reopening in three stages still to come); the notes still say 'final report'. A | · |
Iteration #3 NEEDS_FIXES · 16 findings (truth=7, editorial=6, advisory=3) · Claude Sonnet 5.5 · 30m 59s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F4 hallucinated-fact | · | False against sources the CVE record itself links. cveawg.mitre.org/api/cve/CVE-2026-48356 (dateUpdated 2026-10-28T03:55:44Z, CISA-ADP container, SSVC Exploitation: active) references https://sansec.i | · | |
| F3 claim-not-supported | · | The cited advisory changed after the run's last fetch. https://security.paloaltonetworks.com/json/CVE-2026-109618 timeline (read this iteration): '2026-10-28T05:15:00.000Z | Removed ETAs for 11.2.7-h2 | · | |
| F3 claim-not-supported | · | Same PAN advisory change as above (JSON timeline '2026-10-28T05:15:00.000Z | Removed ETAs for 11.2.7-h23, 11.2.13-h4 and 11.2.16-h1 as these versions are released'); the page now shows 14 released and | · | |
| F3 claim-not-supported | · | https://security.paloaltonetworks.com/json/CVE-2026-111497 timeline: '2026-10-27T05:40 Removed the ETAs for 11.2.10-h17, 12.1.7-h11 and 12.2.4-h1 as these versions are released' and '2026-10-28T05:15: | · | |
| F3 claim-not-supported | · | (low confidence) The JSON timeline carries three entries dated 2026-10-27 (00:10 'Removed the ETAs for 10.2.13-h27 and 10.2.16-h12', 04:03 'Removed the ETAs for 12.1.7-h11 and 12.2.4-h1', 16:40 'Sever | · | |
| F3 claim-not-supported | · | (low confidence) The clause is about what HPE's bulletin says; its cite is the KEV catalog, whose records carry only dateAdded 2026-10-26 and forensicTriage Yes for both CVEs. The absence is verifiabl | · | |
| F3 claim-not-supported | · | (low confidence) Balbooa's page lists 'CVE-2026-102788, CVE-2026-102789' together and two components ('Store & Category Sorting Validation', 'Editor Task Access Control') without saying which CVE is w | · | |
| F9 surface-contradiction | · | (low confidence) Adobe's bulletin APSB26-73 gives 'CVSS base score: 9.6 ... CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:L'; Adobe's own CNA record on cveawg.mitre.org/api/cve/CVE-2026-48356 gives baseS | · | |
| F10 missed-angle | · | The entry's sources are CISA and Adobe only, with verification single-source and credibility 2. The vulnerability has a named, public, mass-exploited history (PolyShell, attacks observed from 2026-03- | · | |
| F8 needs-more-research | · | Sansec (2026-03-17) supplies the specifics the line lacks: exploitation through the anonymous REST guest-cart routes (POST/PUT /V1/guest-carts/:cartId/items), polyglot uploads written to pub/media/cus | · | |
| F5 missing-citation | · | (low confidence) The pattern is asserted with no inline cite in this entry; it is documented in Karimi's honeypot report (four-stage multipart chain, shell parented by an Apache worker) and CISA's PHP | · | |
| F5 missing-citation | · | (low confidence) Same uncited generic pattern as the Cisco entry, and it is stated as what applies to ClearPass nodes although HPE says nothing about ClearPass's PHP deployment (mod_php, FPM or other) | · | |
| F8 needs-more-research | · | (low confidence) The run's trim dropped the source-built tarballs scope: the .NET blog says 'For .NET 10 and .NET 11, CVE-2026-114148 affects source-build tarballs only. For .NET 8 and .NET 9, it also | · | |
| F11 editorial-advisory | · | (low confidence) The type: update record moves updated_at and floats the entry to the top of the live brief at high with an actively-exploited tag, for a delta that is notable-grade (two unauthenticat | · | |
| F11 editorial-advisory | · | Carried over from iterations 1 and 2 (declined, left to the audit). Not raised as a blocker. | · | |
| F11 editorial-advisory | · | Reader-facing provenance should say what the advisory states ('its bug records carry ...') without narrating what the pipeline did not read. | · |
Iteration #4 NEEDS_FIXES · 13 findings (truth=4, editorial=4, advisory=5) · Claude Sonnet 5.5 · 24m 53s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F9 surface-contradiction | · | Only Sansec says 2.4.9 carries the fix ("Update May 12th: Adobe has released Magento 2.4.9 containing the fix"). Adobe's own bulletin APSB26-73 (raw HTML, Date published July 14, 2026) lists "Adobe Co | · | |
| F3 claim-not-supported | · | dotnet/announcements #495 (cited) lists source-built .NET 8 and 9 as affected too: "Source-built .NET 8 Linux distributions | >= 8.0.100, \< 8.0.133 | 8.0.133" and "Source-built .NET 9 Linux distribut | · | |
| F3 claim-not-supported | · | APSB26-73's CVE-2026-48356 row carries "Notes:" with no product mapping, while CVE-2026-47995 carries "Notes: B2B"; the Solution table lists B2B builds for the whole update. Only Adobe's CNA record (n | · | |
| F3 claim-not-supported | · | The Tech Zone post as it stands (Last updated on: October 27, 2026; read through the reader, direct fetches are 403) says "Citrix has released fixes to address these two PHP memory safety vulnerabilit | · | |
| F3 claim-not-supported | · | Uncited, and it reads as if no PAN-OS hotfix has shipped: the advisory (read 2026-10-28 06:18 and 06:36 UTC) carries no ETA on 14 of 20 hotfix builds (released) and ETAs 10/29 or 10/30 on six; the ent | · | |
| F5 missing-citation | · | True (cveawg.mitre.org/api/cve/CVE-2026-48356 CNA container: baseScore 9.3, vector CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N) but the page is neither in sources[] nor cited inline, so a reader cann | · | |
| F5 missing-citation | · | No cite. The supporting page is Inside IT's 2026-09-29 article, which the entry already lists: "Die Cyberkriminellen wollen einen nicht näher spezifizierten Datensatz der Verwaltung gestohlen haben". | · | |
| F17 classification | · | The headline claims added by this run (cause: open-source flaw in the portal's authentication system; no server data left; 71 customers' registration data queried; 206 backups) rest on IDC Frontier's | · | |
| F11 editorial-advisory | · | tools/fetch_source.py extract returns the full 2026-09-29 Inside IT article (trafilatura-direct, 3,161 bytes: Manno isolated the server, SafePay claimed it on its darknet presence, unspecified dataset | · | |
| F11 editorial-advisory | · | git diff HEAD shows the run replaced 'the hotfixes carry estimated dates from 2026-10-27 to 2026-10-30 ([Palo Alto Networks, 2026-10-26])' inside the 2026-10-27 section; the text is right as of 2026-1 | · | |
| F11 editorial-advisory | · | The CSAF note 'Exploitation and Public Discussion' only says the findings 'were originally disclosed in a GitHub security advisory'; it neither reports exploitation nor states there is none. Summary a | · | |
| F11 editorial-advisory | · | The entry does not misstate its delta: summary, record summary, cves[].status ([patch-available]) and the section say plainly that the 2.20.4.2 pair has no reported exploitation. What a reader of the | · | |
| F11 editorial-advisory | · | Declined and left to the audit; not raised as a blocker. | · |
Iteration #5 NEEDS_FIXES · 8 findings (truth=3, editorial=0, advisory=5) · Claude Sonnet 5.5 · 34m 51s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | The sentence ends on the Sansec 2026-03-17 cite, but that page carries neither '2026-jul' nor '2.4.9 included' among the exposed builds; it says 'Update May 12th: Adobe has released Magento 2.4.9 cont | · | |
| F3 claim-not-supported | · | The 'administrator accounts' check is not on the cited 2026-03-17 page (no 'admin' text at all); it is on Sansec's 2026-10-24 post ('Check admin_user for accounts you do not recognize'), which the Det | · | |
| F3 claim-not-supported | · | The cited advisory page (read 2026-10-28 06:51 UTC) carries the hotfix ETAs (six builds due 10/29 or 10/30) but nothing about a Unit 42 retraction; it only links the Unit 42 brief for product protecti | · | |
| F11 editorial-advisory | · | (low confidence) This run's rewrite of the main analysis dropped the paragraph that described Imperva's laboratory demonstration, so the dated 2026-10-26 section now refers to a demonstration the entr | · | |
| F11 editorial-advisory | · | (low confidence, pre-existing) 12 em dashes in reader-facing title, headline and body from earlier fires (check 12 forbids them), and narration such as 'the earlier entry' in the 2026-07-31 section; t | · | |
| F11 editorial-advisory | · | (low confidence, pre-existing) mySites.guru's post says the published CNA record scores CVE-2026-61425 9.4 on a PR:H vector and lists the affected range as '< 1.6.0', and that it keeps its own 10.0 an | · | |
| F11 editorial-advisory | · | Unchanged advisory; left to the audit as the main agent decided. | · | |
| F11 editorial-advisory | · | (low confidence) The supplier's position reaches the entry only through Campingplatz Hüttenberg's notice ('Camping.care untersucht derzeit die Ursache') as relayed by the campsite; the takeaway states | · |
Iteration #6 NEEDS_FIXES · 5 findings (truth=2, editorial=0, advisory=3) · Claude Sonnet 5.5 · 29m 01s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F4 hallucinated-fact | · | Two evidence quotes are no longer contiguous verbatim text on the pages the entry cites, because Citrix revised both on 2026-10-27 and this run kept the old quotes while adding two new ones (the recor | · | |
| F3 claim-not-supported | · | (low confidence) The remediation of the iteration-5 attribution finding added the Huettenberg cite at the end of a clause that also states 'SRF, Blick and a TCS spokesperson tie this theft to the PHP | · | |
| F11 editorial-advisory | · | (low confidence) Sansec's list reads 'Apache pre-2.3.5 without php_flag engine 0' inside a list of Magento versions (next to 'stock nginx 2.0.0-2.2.x'), so 2.3.5 is the Magento release. The entry's ph | · | |
| F11 editorial-advisory | · | (low confidence) The cited 2026-03-17 Sansec page predates any patch ('No official patch exists for production versions') and says only 'Even when execution is blocked, the uploaded file stays on disk | · | |
| F11 editorial-advisory | · | (low confidence, pre-existing) Gridbox: legacy em dashes in title, headline and body, the re-floated card still showing the July headline at priority high with the actively-exploited tag for a delta t | · |
Iteration #7 NEEDS_FIXES · 4 findings (truth=1, editorial=0, advisory=3) · Claude Sonnet 5.5 · 26m 14s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F3 claim-not-supported | · | (low confidence) Body: "Until a build is installed, Citrix tells customers to restrict GUI and API access on those addresses to trusted administrative hosts or narrowly scoped management networks ..." | · | |
| F11 editorial-advisory | · | (low confidence) HPE's bulletin says "customers are asked to consider removing guest portals and all management interfaces from public access until ClearPass instances have been upgraded to a version | · | |
| F11 editorial-advisory | · | (low confidence) The NetScaler entry now quotes the Tech Zone post's 2026-10-27 revision ("Citrix has released fixes to address these two PHP memory safety vulnerabilities for all supported NetScaler | · | |
| F11 editorial-advisory | · | (low confidence) Pre-existing wording and structure from earlier fires that this run did not add; the 2026-10-28 Gridbox section itself says plainly that nothing in it is reported exploited and the en | · |
Iteration #8 CLEAN · 2 findings (truth=0, editorial=0, advisory=2) · Claude Sonnet 5.5 · 30m 55s
| F-code | Section | Item · URL/quote | Verifier summary | Remediation · outcome |
|---|---|---|---|---|
| F11 editorial-advisory | · | (low confidence) The update relabelled the Inside IT 2026-09-29 source as 'article of 2026-09-29' and the page now reads in full (it also carries the SafePay claim of an unspecified dataset), but the | · | |
| F11 editorial-advisory | · | (low confidence) The Oracle entry states 90 unauthenticated network rows at 9.0 or higher covering 73 distinct CVE ids ('90 of those, covering 73 distinct CVE ids'); the note's unit is wrong (73 ids, | · |
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
2026-10-28T0256Z-intel · Sonnet 5.5 · window 26 h · 5 entries published
Verification & coverage notes
Window 26 h (gap 24.0 h to 2026-10-27T0256Z-intel, normal cadence, no outage backfill). entities/registry.yaml (672 KB) was grepped per key instead of read whole; the 14-day prior_coverage.json (70 records) was read in full. Slice build: S1 and S2 excluded 0 sources by the two-previous-fires rule, S3 excluded 1, S4 took its whole 14-source pool. No intel drops, so no S5. Every slice ledger row came back attempted (17, 22, 14 and 14 records); no continuation was needed.
Verification. Iterations 1 to 7 each returned NEEDS_FIXES (their findings and remediations are in the record above) and the first CLEAN landed at the 8-iteration cap, so no confirmation pass was possible: the run publishes on that single CLEAN as a fail-open and the waiver is recorded. Open advisories, all pre-existing wording or length from earlier fires and left to the quality audit: the Gridbox entry's legacy em dashes, its re-floated high priority and the CVE-2026-61425 CNA score difference (9.4 against the discloser's 10.0), and the PHP entry's length on a non-deep-dive.
What this fire changed. Five new entries (Cisco PHP2Shell advisory, HPE ClearPass PHP2Shell, the CISA-listed Adobe Commerce upload flaw CVE-2026-48356, ARA-Region Lyss-Limpachtal, DeadTag) and nine changelog records. The two that matter most: Palo Alto's advisory of 2026-10-27 now states active exploitation against internet-exposed PAN-OS management interfaces (PAN-OS entry notable to high, PHP entry stays critical with the statement added), and CISA's KEV listing of the Brotli decoder flaw CVE-2026-111497 on 2026-10-27 moved the Node2Shell entry to exploited and high.
KEV rows (kev-window.txt).
- CVE-2026-111497: COVERED-STALE on two entries, both given
updaterecords (Node2Shell entry; the .NET entry, whose CVE record now carriesexploited,cisa-kevand says Microsoft's own record still read not exploited on 2026-10-25). - CVE-2026-48356 (Adobe Commerce and Magento upload flaw, KEV 2026-10-27): NOT COVERED, shipped as a notable entry. It was first held back as borderline (CISA's listing and Adobe's bulletin give no exploitation detail, no constituency nexus); the verifier then found Sansec's PolyShell reporting, which the CVE record itself links: attacks since 2026-03-19, a mass wave on 2026-03-30 and the fix timeline (2.4.9 on 2026-05-12, supported-line backport 2026-07-14). The entry now rests on Sansec, Adobe and CISA together, and the store already carries four Adobe Commerce or Magento entries.
- No RANSOMWARE rows.
Backlog (state/coverage_backlog.md). ARX Group SA (Aurora leak-site claim): held, unchanged: S2 and S4 searched in German, Italian and English and read Inside IT, Netzwoche, SRF, Blick, ICTjournal, NZZ, RTS and swisscybersecurity.net; no ARX, SBB or ASTRA/USTRA statement and no Swiss press report exist, expiry 2026-11-07. The 2026-10-14 strike of the ARA Lyss row is overtaken by Inside IT's report of 2026-10-27 and the new ARA Lyss entry.
Supersessions and corrections. Palo Alto's statement replaces the PAN-OS entry's "no malicious exploitation" text and its availability-only score (8.7 to 9.2); the PHP entry's "no source ties a named CVE to a production compromise" now carries the Palo Alto exception. Camping.Care: a second customer's notice says the supplier is investigating the cause and does not mention PHP, so the entry now says the PHP link rests on SRF, Blick and TCS. Manno: the entry's "claim could not be corroborated, no dataset" statements are replaced by Inside IT's report that about 30 GB were published (single source, Risikomonitor's analysis unverified, rating B3). IDC Frontier: the seventh report names the cause, so "intrusion route not public" is replaced.
Contradictions and borderline includes.
- Contradiction: Palo Alto's exploitation wording. An Internet Archive copy of the advisory taken 2026-10-27 17:10 UTC says "very limited exploitation", the page read 2026-10-28 says "active exploitation ... exposed to the internet", and the vendor timeline records no such change; the PAN-OS and PHP entries state both and tie neither to the 16:40 UTC revision.
- Contradiction: Adobe's bulletin APSB26-73 (2026-07-14, updated 2026-08-11) says no exploits are known for CVE-2026-48356, while Sansec reports attacks since 2026-03-19 and CISA lists the flaw as exploited; Adobe's CVE record scores it 9.3 where the bulletin says 9.6. The entry states both.
- Contradiction: HPE HPESBNW05171 names ClearPass 6.14.1 and 6.11.16 as versions "providing some level of mitigation" in its workaround text, while its affected list covers 6.14.1 and 6.11.16 and below and its remediation names hotfix patches for both; the entry quotes the bulletin and does not resolve it.
- Borderline include: Balbooa Gridbox CVE-2026-102788 and CVE-2026-102789, an unauthenticated read-side SQL injection with no exploitation report, folded into the July Gridbox entry as an
updaterecord (the store's pattern for later Gridbox batches, which moved its action to 2.20.4.2). Kept under the "otherwise" limb of PD-11 (b), the leg that holds being a two-file fix diff that hands over the technique; the July Gridbox flaws were exploited, these two are not reported exploited, and no Swiss nexus is established.
Single-source and reduced-confidence inclusions.
- ARA Lyss-Limpachtal: Inside IT relays Risikomonitor's analysis; DeXpose carries the claim; C3, routine, no encryption mapped because no source states it.
- Cisco, HPE ClearPass, Balbooa (with mySites.guru, on the Gridbox update): vendor first-party; Cisco and HPE carry
cvss: n/aor the vendor's score only. - Adobe Commerce CVE-2026-48356: the exploitation history rests on Sansec's own telemetry and scans (a security vendor) with CISA's listing and Adobe's bulletin; confidence high, reliability B.
- DeadTag: Report URI and iVerify published jointly, so one account from two vantage points, credibility 2.
borderline-drop lines.
- Gotenberg CVE-2026-42589 (VulnCheck canary telemetry, 1,358 events): internal document-conversion service, five-month-old flaw, not in CISA KEV, no established constituency exposure.
- Oracle October CPU, 90 unauthenticated rows at CVSS 9.0 or higher (73 distinct CVE ids) against the 12 in the existing entry's title: the entry states the full tally, the CPU is applied as a whole and none is reported exploited; not widened. Oracle Alert CVE-2026-109618 rev 2 Oracle Communications rows: telecom signalling products, no nexus (Secure Backup and Solaris are already in the OT-products entry).
- Cisco eMudhra informational (revised 2026-10-27): vendor-status churn on a covered entry that already carries the Cisco rows.
- Chrome 156.0.8078.73/.74 (52 fixes, four Critical), Mozilla MFSA 2026-118 to 122, Apple iOS 27.2, 26.7.2 and 18.7.11 (162 CVEs), GLPI 11.0.9 and 10.0.27, Joomla extension releases of 2026-10-26/27, EUVD criticals (Pie Register, llama.cpp ggml-rpc, OpenRemote): none reported exploited, no constituency nexus.
- Tailscale and headscale audit by Cyberdione Labs ("Tailworm", fixed in v1.104.4): needs an approved peer whose policy allows forwarding, no CVE, no exploitation, constituency exposure not established.
- Yoba (Check Point, fake GitHub AI-app projects with DLL sideloading): commodity technique, no constituency specificity.
- Arctic Wolf on Microsoft Teams external-chat impersonation: known technique, primary dated 2026-10-26 outside the window, Swiss relay only.
- HMICFRS inspection of police and fire and rescue cyber posture: UK regulator report, awareness level, no Swiss obligation.
- Kaspersky managed-Kubernetes isolation chain (provider anonymised, fixed), EDK II CVE-2026-97631 (the 2026-10-25 audit's reopen conditions are unmet), Popcorn Time updater (sinkholed since March, no exploitation), Sysdig pickle research, Bishop Fox PHP2Shell guidance (marginal delta on an entry that already carries KEV, public exploit and hardening): no change to what a Swiss SOC does.
- Swiss Ice Hockey Federation MYSIHF (valid-account theft, private federation), BACS weekly review week 43 (WhatsApp account takeover, consumer fraud; week 42 was also not published): no public-sector nexus.
- Moldova government digital services (about 750,000 records through unauthenticated services, no service or flaw named), University of Copenhagen test environment, GMO Onamae Desktop Cloud (Japan), Navigate360 / Canadian Crime Stoppers (investigation opened by the Privacy Commissioner), UNM Health, ENGIE, Medienwerkstatt Sulgen: out of nexus or no vector, actor or behaviour beyond the impact. Leak-site claims without victim or A/B-press confirmation (ECOM, SunTechnics Fabrisolar, Bristol City Council, OCMW Machelen, Alliance Police Nationale) stay out (PD-6).
- Convertr / Plausible CE CVE-2026-8467 (Magecart skimmer, suspected entry): suspected only, out of nexus; a pointer for the next audit.
Sources. karimi-honeypot-data promoted to active (promotion_due). Two candidates added because published entries cite them: ellio-threat-research, jfrog-security-research. Not added: Eye Security Research and Eclypsium (S3 leads, unused), risikomonitor-news and hacksnack.ch (S2 and S4 leads, reliability C and D, relays only). Recipe notes recorded for hpe-networking-psirt (changes.csv), cisco-psirt (feed), paloalto-psirt (advisory JSON), inside-it-ch (__NEXT_DATA__), apa-ots, ransomware-live and databreaches-net. connectwise-trust-center and ec-digital-strategy-newsroom stay 403 on every transport; a 403 never demotes.
Registry. Added incident:ara-region-lyss-limpachtal-ransomware-2026-09, campaign:deadtag, tool:coruna and tool:darksword, the aliases Brotli2Shell, HTTP2Shell and Broccoli2Shell on tool:node2shell, and seven product entities through sync_products.py (five from the new entries, Bun and Istio from the Node2Shell update).
Watchlist: products checked=0, hits=0; suppliers checked=0, hits=0 (no watchlists configured).
Coverage gaps: connectwise-trust-center (403 on every transport); ec-digital-strategy-newsroom (403 on every transport, third run); resecurity (empty JavaScript shell, newest post 2026-10-14 by search); golem-security (cookie wall); apa-ots (feed holds only the 15 newest releases, no lookback sweep possible); nau-ch, push-security, novee-security (nothing in the horizon; novee-security marketing only, demotion candidate); Juniper PHP2Shell reference advisory (Salesforce shell on every reader) and QNAP QSA-26-18 (reader returned navigation only) were not re-read, so any QNAP revision after 2026-10-27 is unverified; the Citrix community post is 403 on every direct transport and was read only through the reader.
← Operations dashboard · run-record contract: docs/pipeline.md