Add the seven OpenClaw findings from the AI SAST campaign

[p80n-sec]

Adds the OpenClaw campaign: six findings covered by the AI SAST data-flow post, plus the apply_patch path traversal disclosed two days later with its own writeup. Both articles were added to links.yml.

Record	Key	CVSS	Disclosed
Missing Telnyx webhook auth	CVE-2026-26319	7.5	2026-02-17
Gateway tool gatewayUrl SSRF	CVE-2026-26322	7.6	2026-02-17
Image tool remote-fetch SSRF	GHSA-56f2-hvwg-5743	7.6	2026-02-17
Tlon (Urbit) auth SSRF	CVE-2026-28476	6.3	2026-02-18
Twilio webhook bypass via ngrok loopback	CVE-2026-29606	6.3	2026-02-18
Browser upload path traversal	CVE-2026-26329	7.1	2026-02-18
apply_patch path traversal	CVE-2026-32060	8.7	2026-02-19

Unrelated to the DEF CON 34 orchestration research, so none of these carry a talk. All seven credit Peyton Kennedy alone, which matches the sole reporter on every advisory. disclosed is each advisory's publication date; all seven were reported 2026-02-04.

Identifiers follow the usual precedence — the CVE where one exists, and GHSA-56f2-hvwg-5743 for the image-tool SSRF, which never got one.

On the scores

The number recorded is the one each advisory publishes as its headline severity: CVSS v4.0 where a v4 score exists, v3.1 otherwise. That is what the source table quotes, and it is what GitHub displays. It does mean three records carry a number lower than their own v3.1 score, so each says so in its body:

Record	Recorded (v4.0)	v3.1 on the same advisory
CVE-2026-28476	6.3	8.3
CVE-2026-29606	6.3	6.5
CVE-2026-32060	8.7	8.8

CVE-2026-26329 had no score at all when its advisory published — the article says "High (No CVSS Assigned)" — and now carries a v4.0 score of 7.1 and no v3.1 score, so 7.1 is the only number available.

Worth noting this differs from the DEF CON records, which are all v3.1 because that is what the whitepaper and the source table quoted there.

One discrepancy in the article

The AI SAST post's summary table lists GHSA-4hg8-92x6-h2f3 twice, on both the Telnyx row and the Urbit row. The Urbit finding is GHSA-pg2v-8xwh-qhcc; the records use the correct ids. The same table also predates two CVE assignments, listing Urbit and Twilio as "No CVE ID" where CVE-2026-28476 and CVE-2026-29606 now exist.

Verification

.venv/bin/python -m pytest -q                                   # 176 passed
.venv/bin/python -m tools.generate --today 2026-08-24
.venv/bin/python -m tools.generate --check --today 2026-08-24   # exit 0

All 26 records render and every detail page builds. github.com/advisories/GHSA-56f2-hvwg-5743 resolves, so the one GHSA-keyed record links live.

🤖 Generated with Claude Code