Status: Owner decision (Emil, 2026-11-08). TFP-274 done (e9ab8b5); TFP-273 done (99b477c). TFP-285 done (f48579f, f354687).
Emil asked for a way to see dependency issues and clear them with patches, not only dismiss them. Two parts, both chosen on 2026-11-08:
Each item says what is waiting, what would end the wait, and how to check. The monthly check reads this section; add an item whenever a fix is held back, and remove it (with the date and commit) when it is cleared.
| Item | Waiting on | Check | Then |
|---|---|---|---|
| W1. minimatch below 11.0.1 (GHSA-g3hv-75wg-m249; Dependabot #144 and #145 dismissed as tolerable risk, 2026-11-04 and 2026-11-08) | ESLint’s plugins (eslint-plugin-import, jsx-a11y, react) accepting minimatch 11, which has no default export; Allure’s path too |
npm view eslint-plugin-import dependencies.minimatch, the same for the other two and for @allurereport/ci; any new Dependabot alert on minimatch |
Upgrade, drop any override, reopen nothing |
W2. The DOMPurify override (package.json overrides.dompurify, TFP-270) |
An Allure release whose @allurereport/web-commons depends on dompurify 3.4.17 or later |
npm view @allurereport/web-commons dependencies.dompurify |
Upgrade Allure, remove the override and its "//" note |
| W3. Next.js (16.4.3, pinned) | Patch and minor releases, security advisories; AGENTS.md says read the bundled docs before changing code |
npm view next version; the Next.js security advisories |
A ticket per upgrade; majors get a decision record |
W4. Playwright (@playwright/test ^1.59.1, Chromium only) |
New minors (browser updates, fixes) | npm view @playwright/test version |
A ticket; the template check and regression suites must stay green |
W5. Node (24, .nvmrc) |
Node 24 patch releases and security releases; the next LTS | nodejs.org release schedule and security blog | Update .nvmrc and CI in a ticket |
| W6. Open alerts | Anything open in Dependabot, code scanning or secret scanning | The repo’s Security tab | One line each, with severity and a suggested ticket |
| W7. The nightly’s audit | npm audit with dev dependencies included (the nightly only checks production) |
npm audit and npm audit --omit=dev on main |
New high or critical findings become tickets |
99b477c, verified (Cowork, 2026-11-08).
First run (13:48Z, both update jobs succeeded): npm opened 5 pull
requests, #18 the nextjs group (4), #19 npm-minor-and-patch (10),
#20 openai 7, #21 typescript 7, #22 eslint 10; the limit held back none;
github-actions, none (every action on its latest major). CI: #18, #19,
#20 and #22 green; #21 fails lint (typescript-eslint does not support
TS 7.0). .github/dependabot.yml (npm and github-actions, monthly,
limit 5; npm minors and patches in one group, majors one per package;
next, eslint-config-next, @playwright/test and playwright-core
majors ignored) and a merge rule in docs/QUALITY_GATES.md §3. Read
by Cowork: the config does what the scope says; the merge rule’s
reason for not using the template check (both refs share one
node_modules) is sound, so the unit and regression suites stand in for
it. Claude Code’s schema check (valid; a broken copy invalid, naming both
fields) wasn’t re-run here (SchemaStore isn’t reachable from Cowork’s
workspace). Rulings (Emil, 2026-11-08): (1) next, react,
react-dom and eslint-config-next in a group of their own, so a
Next.js update arrives as its own pull request (W3); (2) @types/node
majors ignored while .nvmrc is 24 (W5); (3) the first run’s pull
requests reported, with how many the limit held back; (4) Dependabot
security updates turned on in the repository’s settings (by Cowork), so
an alert with a patch gets a fix pull request (a minimatch 11 one will
fail CI and be closed, W1); the config’s header says so. All four are
in the commit.ci, dependencies.github/dependabot.yml with two ecosystems, npm (root) and
github-actions, each on a monthly schedule.next majors ignored (a major needs a decision record);
@playwright/test kept to minors in the group.docs/QUALITY_GATES.md or CONTRIBUTING.md says how a Dependabot pull
request is merged (CI green, template check identical or explained).e9ab8b5, verified (Cowork, 2026-11-08);
#6133 closed as fixed by CodeQL’s scan of e9ab8b5 (13:24:46Z;
JavaScript results 16 to 15). shouldQueueUrl queues only http: and https:
(CRAWLED_PROTOCOLS), exported for its test. Verified on a Linux
clone, the uncommitted files on 59eb2fd: crawl.test.ts and the crawl
route test 5 of 5 passed, 0 skipped; with the old blocklist (exported),
the scheme-only test fails on vbscript:msgbox(1), so this ticket’s
“can’t fail on today’s code” holds only for the origin-level test.
Claude Code’s template check (172 identical) and unit run (769 of 769)
weren’t re-run here. #6133 closing is checked after the push.main)security, crawlb493265’s move) flags shouldQueueUrl in src/lib/pipeline/crawl.ts
(line 162 at 6b617b2): it refuses mailto:, tel:, javascript: and
data: by a blocklist, which misses vbscript: and anything else. The
origin check just before it already refuses every non-http(s) URL, so
it isn’t reachable today, but a blocklist is the wrong shape.http:
and https: URLs on the crawl’s origin.http: and https: URLs on the crawl’s origin
are queued (vbscript: and an unknown scheme are not). It can’t fail
on today’s code, because the origin check already refuses those
schemes; the commit says so (docs/QUALITY_GATES.md §1).check:template-unchanged identical; the crawl tests pass unchanged.f48579f (the lockfile), f354687 (one
check:old-name allowlist entry for the old command’s bin line, which
npm copied into the lockfile; the owner chose this, 2026-11-12), on
main by fast-forward (PR #29). package-lock.json has one
fast-xml-parser, 5.12.0. After npm ci: test:unit 821 of 821,
0 skipped; test:self:allure 5 passed, its Allure report 5 of 5;
test:regression:allure 73 passed, its report 73 of 73. PR
Validation and CodeQL green on f354687. The dependency graph
shows 5.12.0, and Dependabot closed its own bump PR #27 at 16:19Z.
Alert #146 was still open at 16:28Z: it closes on Dependabot’s next
scan of main. Added 2026-11-12 (owner decision: patch it).main)security, dependenciesmain:
fast-xml-parser up to 5.11.2 rescans the rest of the input from every
< inside a stop node that is never closed, so a crafted input costs
O(n²) CPU before the parser throws. It needs the non-default stopNodes
option. The advisory calls it a static-analysis finding, not
benchmarked. Here it is reached only through Allure, a dev dependency:
allure → @allurereport/core → @allurereport/reader →
fast-xml-parser 5.11.2. The reader parses our own test results when
test:self:allure and test:regression:allure build their reports,
and none of its four parsers sets stopNodes. Nothing in src/,
scripts/ or bin/ uses the package.npm update fast-xml-parser. The reader asks for ^5.11.1,
so 5.12.0 is within range: only package-lock.json changes, with no
override and no Allure upgrade.docs/QUALITY_GATES.md §3, a dependency
change):
package-lock.json changes; npm ls fast-xml-parser shows
5.12.0 or later.npm ci: npm run test:self:allure and
npm run test:regression:allure pass and write their Allure reports;
npm run test:unit passes. The counts are in the commit body.main, checked after the push.