fix(security): bump @hono/node-server to ^2.0.12 (GHSA-frvp-7c67-39w9)

CHANGELOG.md

@@ -32,6 +32,36 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Documented Kubernetes/IaC as follow-up work rather than a current
   blocker for this static app.
 
+### Security
+
+- Removed the dynamic `new RegExp(...)` in `parseMsProjectXml`'s `tag()` helper
+  (`cloud-sync.js`) that the SAST gate flagged for potential ReDoS
+  (`detect-non-literal-regexp`). Every caller passes a hardcoded tag name and
+  the pattern was linear, so it was not exploitable, but the extractor now uses
+  `indexOf` slicing with identical semantics — clearing the finding at base
+  rather than suppressing it. Verified unchanged by the MS Project import unit
+  tests.
+- Upgraded the `@hono/node-server` runtime dependency from `^1.19.14` to
+  `^2.0.12` to clear GHSA-frvp-7c67-39w9 (moderate: `serve-static` path
+  traversal on Windows via an encoded backslash). ScopeWeave only calls the
+  stable `serve()` entry — it does not use `@hono/node-server`'s `serveStatic`
+  helper, so the vulnerable path was unreachable, but the dependency is bumped
+  to keep the lockfiles clean. Verified with `npm run test:unit` and
+  `npm run test:api` (server boots and passes the API + rate-limit smoke under
+  v2); `package-lock.json` updated.
+
+### Removed
+
+- Removed the vestigial `pnpm-lock.yaml`. ScopeWeave is an npm project
+  (`package-lock.json` is canonical; `server-tests.yml`/`fuzz.yml` use
+  `cache: 'npm'` + `npm ci`; no workflow, Dockerfile, or compose file
+  references pnpm), so the committed pnpm lock was a second, unused lockfile.
+  The central OpenCode coverage-evidence runner selects pnpm whenever a
+  `pnpm-lock.yaml` is present and then refuses because `package.json` declares
+  no exact `packageManager: pnpm@X.Y.Z`, which blocked coverage evidence — and
+  therefore review approval — on every ScopeWeave PR. Deleting the vestigial
+  lock lets the runner use the canonical npm path.
+
 ## [1.0.0] - 2026-04-20
 
 <!-- markdownlint-disable-next-line MD024 -->

cloud-sync.js

@@ -739,9 +739,20 @@ function openReportModal() {
 // no DOMParser needed → node-testable); swap for a real XML parser if
 // hand-edited files ever matter.
 export function parseMsProjectXml(xml) {
+  // Extract the text of a simple <name>…</name> element without building a
+  // dynamic RegExp from `name` (which trips SAST ReDoS/regex-injection rules,
+  // even though every caller passes a hardcoded tag). indexOf slicing mirrors
+  // the original /<name>([^<]*)<\/name>/ semantics: the run must contain no
+  // '<', so a nested tag yields '' rather than a false capture.
   const tag = (block, name) => {
-    const m = block.match(new RegExp(`<${name}>([^<]*)</${name}>`));
-    return m ? m[1].trim() : '';
+    const open = `<${name}>`;
+    const start = block.indexOf(open);
+    if (start === -1) return '';
+    const from = start + open.length;
+    const end = block.indexOf(`</${name}>`, from);
+    if (end === -1) return '';
+    const content = block.slice(from, end);
+    return content.includes('<') ? '' : content.trim();
   };
   const unescape = (s) => s
     .replace(/&lt;/g, '<').replace(/&gt;/g, '>').replace(/&quot;/g, '"')

package.json

@@ -17,7 +17,7 @@
     "fuzz": "node --test tests/fuzz/*.mjs"
   },
   "dependencies": {
-    "@hono/node-server": "^1.19.14",
+    "@hono/node-server": "^2.0.12",
     "hono": "^4.12.27"
   },
   "devDependencies": {

tests/unit/msproject.test.mjs

@@ -55,4 +55,33 @@ assert.equal(t5.predecessors, 'msp-1');
 
 assert.deepEqual(parseMsProjectXml('<Project></Project>'), [], 'no tasks → empty');
 
+// --- tag() failure-branch regression tests (indexOf-slicing extractor) ---
+// These pin the three defensive branches of the `tag()` helper introduced when
+// the dynamic RegExp was removed, so a future parser change cannot silently
+// alter how malformed or nested markup is handled.
+
+// end === -1: <Name> opens but never closes → '' → task has no usable name → skipped.
+assert.deepEqual(
+  parseMsProjectXml('<Task><UID>10</UID><Name>broken</Task>'),
+  [],
+  'unterminated <Name> (end === -1) yields no name → task skipped',
+);
+
+// content.includes('<'): nested markup inside the value → '' → task skipped
+// (mirrors the original /<name>([^<]*)<\/name>/ "no inner <" semantics).
+assert.deepEqual(
+  parseMsProjectXml('<Task><UID>11</UID><Name>a<b>c</b></Name><OutlineLevel>1</OutlineLevel></Task>'),
+  [],
+  'nested markup inside <Name> (content contains "<") is rejected → task skipped',
+);
+
+// start === -1: absent optional tags → '' (dates blank, OutlineLevel defaults to 1).
+const sparseTasks = parseMsProjectXml('<Task><UID>12</UID><Name>NoDates</Name></Task>');
+assert.equal(sparseTasks.length, 1, 'valid UID + Name parses even with no dates/level');
+assert.equal(sparseTasks[0].id, 'msp-12');
+assert.equal(sparseTasks[0].depth, 1, 'missing <OutlineLevel> (start === -1) defaults depth to 1');
+assert.equal(sparseTasks[0].plannedStartDate, '', 'missing <Start> (start === -1) → empty date');
+assert.equal(sparseTasks[0].plannedEndDate, '', 'missing <Finish> (start === -1) → empty date');
+assert.equal(sparseTasks[0].actualProgress, 0, 'missing <PercentComplete> → 0');
+
 console.log('✓ MS Project import tests passed');