Add pre-commit hook and github action for secret scanning

.github/workflows/secret-scan.yml

@@ -0,0 +1,32 @@
+name: secret-and-phi-scan
+
+# Server-side backstop for the local pre-commit hook (which a developer can skip with
+# --no-verify). Runs the same .gitleaks.toml ruleset against the checked-out tree.
+#
+# We invoke the gitleaks *binary* via its Docker image rather than gitleaks/gitleaks-action,
+# because that Action requires a GITLEAKS_LICENSE for organization-owned repos. The binary
+# itself is MIT-licensed with no such requirement.
+
+on:
+  pull_request:
+  push:
+    branches: [main]
+
+permissions:
+  contents: read
+
+jobs:
+  gitleaks:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Scan working tree for secrets + PHI
+        run: |
+          docker run --rm -v "${{ github.workspace }}:/repo" \
+            ghcr.io/gitleaks/gitleaks:v8.30.1 \
+            dir /repo \
+            --config /repo/.gitleaks.toml \
+            --redact \
+            --verbose \
+            --exit-code 1

.gitleaks.toml

@@ -0,0 +1,62 @@
+# gitleaks config — shared by the pre-commit hook and the CI scan.
+# Goal: keep secrets and obvious PHI/PII out of GitHub (see threat model — we protect the
+# remote, not LLM egress). Built-in secret rules are inherited; the custom rules below add
+# high-signal PHI/PII patterns. Deliberately NOT flagged: bare NPIs and dates of birth — on
+# their own those are not PHI violations for this repo, and regexing them is pure noise.
+#
+# Free-text patient NAMES are intentionally absent: there is no low-false-positive regex for
+# them. Names stay a code-review concern (especially for hand-authored queries/*.sql).
+title = "Chamber Cardio nao-analytics-agent — secrets + PHI/PII gate"
+
+[extend]
+# Inherit all of gitleaks' built-in secret detectors (Anthropic sk-ant- keys, AWS keys,
+# private keys, generic high-entropy tokens, etc.).
+useDefault = true
+
+# --- Global allowlist ----------------------------------------------------------------------
+[[allowlists]]
+description = "Paths that are machine-generated or hold no patient data."
+paths = [
+  '''poetry\.lock''',
+  '''\.gitleaks\.toml''',   # don't flag the example patterns in this file itself
+]
+
+# --- Secret rules (augment the built-ins) --------------------------------------------------
+
+[[rules]]
+id = "anthropic-api-key"
+description = "Anthropic API key (sk-ant-...). Our #1 secret; gitleaks' defaults miss it."
+regex = '''sk-ant-[A-Za-z0-9]+-[A-Za-z0-9_-]{20,}'''
+keywords = ["sk-ant-"]
+tags = ["secret", "anthropic"]
+
+# --- PHI / PII rules -----------------------------------------------------------------------
+
+[[rules]]
+id = "phi-us-ssn"
+description = "Possible US Social Security Number (PHI)"
+regex = '''\b\d{3}-\d{2}-\d{4}\b'''
+tags = ["PHI", "PII"]
+
+[[rules]]
+id = "phi-us-phone"
+description = "Possible US phone number (PII)"
+# Requires delimiters between groups so a bare 10-digit NPI (e.g. 1234567890) does NOT match.
+regex = '''\b(?:\+?1[-.\s])?\(?\d{3}\)?[-.\s]\d{3}[-.\s]\d{4}\b'''
+tags = ["PII"]
+
+[[rules]]
+id = "phi-email"
+description = "Email address (possible patient PII)"
+regex = '''[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}'''
+tags = ["PII"]
+
+  [[rules.allowlists]]
+  description = "Known-safe, non-patient addresses and infra hostnames."
+  regexTarget = "match"
+  regexes = [
+    '''@chambercardio\.com''',          # our own corporate domain
+    '''git@github\.com''',              # SSH remote in nao_config.yaml / sync script
+    '''@(example|test|invalid)\.(com|org|net)''',
+    '''noreply@''',
+  ]

.pre-commit-config.yaml

@@ -0,0 +1,8 @@
+default_stages: [pre-commit]
+repos:
+  - repo: https://github.com/gitleaks/gitleaks
+    rev: v8.30.1
+    hooks:
+      - id: gitleaks
+        name: 🔐 secrets + PHI scan
+        description: "Block secrets and obvious PHI/PII from being committed (uses .gitleaks.toml)"