ci: check fixture identifiers by allowlist instead of denylist

.github/workflows/ci.yml

@@ -53,15 +53,8 @@ jobs:
             echo "expected at least 17 golden vectors, found $count" >&2
             exit 1
           fi
-      - name: Fixtures must not carry customer or internal identifiers
-        run: |
-          set -e
-          # This repository is public. Fixtures use neutral identifiers only.
-          if grep -rniE 'ajj|finops-core|debtor_aging|autocount' testdata README.md src; then
-            echo "found a real customer or internal identifier in public files" >&2
-            exit 1
-          fi
-          echo "no real identifiers present"
+      - name: Fixture identifiers must be neutral
+        run: node scripts/check-fixture-identifiers.mjs
 
   sast:
     name: SAST (CodeQL)
@@ -80,10 +73,21 @@ jobs:
   secret-scan:
     name: Secret scan
     runs-on: ubuntu-latest
+    env:
+      # The gitleaks GitHub Action requires a paid licence for organisation
+      # repositories; the CLI it wraps does not. Version pinned deliberately.
+      GITLEAKS_VERSION: 8.28.0
     steps:
       - uses: actions/checkout@v4
         with:
           fetch-depth: 0
-      - uses: gitleaks/gitleaks-action@v2
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      - name: Install gitleaks
+        run: |
+          set -euo pipefail
+          curl -sSfL "https://github.com/gitleaks/gitleaks/releases/download/v${GITLEAKS_VERSION}/gitleaks_${GITLEAKS_VERSION}_linux_x64.tar.gz" \
+            | tar -xz gitleaks
+          ./gitleaks version
+      - name: Scan working tree
+        run: ./gitleaks dir . --no-banner --redact
+      - name: Scan git history
+        run: ./gitleaks git . --no-banner --redact

scripts/check-fixture-identifiers.mjs

@@ -0,0 +1,93 @@
+// Fixture identifiers must stay neutral: this repository is public.
+//
+// An allowlist is used rather than a list of forbidden strings. A denylist would
+// have to name the very identifiers that must not appear here, and it only ever
+// catches the names someone remembered to add. The allowlist catches any
+// unexpected identifier, including ones introduced later.
+import { readFileSync, readdirSync } from 'node:fs';
+import { join } from 'node:path';
+
+const TENANT_OK = /^(test-tenant|tenant-[a-z0-9-]+)$/;
+const AGENT_OK = /^agent-[a-z0-9-]+$/;
+// Kid may carry dots and underscores because one vector pins kid quoting. Written as a
+// single character class rather than a repeated group: with '-' in both a group body and
+// its separator class, the input can be partitioned ambiguously and the match backtracks
+// exponentially.
+const KID_OK = /^agent-[a-z0-9._-]+$/;
+
+// Identifiers appear both as JSON fields and inside HTTP header values, so both
+// shapes are inspected. A field-only check missed a planted name in the header set.
+const FIELD_RULES = [
+  ['tenant_id', TENANT_OK],
+  ['agent_id', AGENT_OK],
+  ['kid', KID_OK],
+  ['X-Openmax-Tenant', TENANT_OK],
+  ['X-Openmax-Agent', AGENT_OK],
+];
+
+// keyid="..." inside Signature-Input carries the kid.
+const KEYID_IN_SIGNATURE_INPUT = /keyid="([^"]+)"/g;
+
+const violations = [];
+
+function walk(value, path) {
+  if (Array.isArray(value)) {
+    value.forEach((v, i) => walk(v, `${path}[${i}]`));
+    return;
+  }
+  if (value === null || typeof value !== 'object') return;
+  for (const [key, v] of Object.entries(value)) {
+    for (const [field, ok] of FIELD_RULES) {
+      if (key === field && typeof v === 'string' && !ok.test(v)) {
+        violations.push(`${path}.${key} = ${JSON.stringify(v)} (expected ${ok})`);
+      }
+    }
+    if (key === 'Signature-Input' && typeof v === 'string') {
+      for (const m of v.matchAll(KEYID_IN_SIGNATURE_INPUT)) {
+        if (!KID_OK.test(m[1])) {
+          violations.push(`${path}.${key} keyid = ${JSON.stringify(m[1])} (expected ${KID_OK})`);
+        }
+      }
+    }
+    walk(v, `${path}.${key}`);
+  }
+}
+
+const dir = 'testdata';
+const files = readdirSync(dir).filter((f) => f.endsWith('.json'));
+if (files.length === 0) {
+  console.error('no fixture files found in testdata/ — the check would pass vacuously');
+  process.exit(1);
+}
+
+let inspected = 0;
+for (const f of files) {
+  const parsed = JSON.parse(readFileSync(join(dir, f), 'utf8'));
+  walk(parsed, f);
+  inspected += 1;
+}
+
+// Guard against a rule set that matches nothing: at least one value per field must exist.
+const seen = new Set();
+function collect(value) {
+  if (Array.isArray(value)) return value.forEach(collect);
+  if (value === null || typeof value !== 'object') return;
+  for (const [key, v] of Object.entries(value)) {
+    if (FIELD_RULES.some(([field]) => field === key) && typeof v === 'string') seen.add(key);
+    collect(v);
+  }
+}
+for (const f of files) collect(JSON.parse(readFileSync(join(dir, f), 'utf8')));
+for (const [field] of FIELD_RULES) {
+  if (!seen.has(field)) {
+    console.error(`no ${field} values found in any fixture — the check is not actually inspecting anything`);
+    process.exit(1);
+  }
+}
+
+if (violations.length > 0) {
+  console.error('non-neutral identifiers in public fixtures:');
+  for (const v of violations) console.error(`  ${v}`);
+  process.exit(1);
+}
+console.log(`fixture identifiers neutral (${inspected} file(s), fields checked: ${[...seen].join(', ')})`);