Files
VoiceStudio/.github/workflows/security.yml
T
Palash DebnathandClaude Opus 4.8 4ec5b4cb7d chore(security): scope CodeQL to shipped product code (#201)
CodeQL flagged 459 alerts on main, but triage showed the bulk are in code
that never ships in the installer's runtime path: file-not-closed in the
omnivoice/eval harnesses, unused-global "FPs" in alembic migration boilerplate
(revision/down_revision), bind-all in tests, and path sinks in the legacy
Gradio research UI. They drowned out the handful of real findings.

Add a CodeQL config (inline, supported because build-mode is `none`/interpreted)
that scopes analysis to product code via paths-ignore: omnivoice/eval, research,
tests, backend/migrations, and *.test.* files. Queries move into the inline
config so security-and-quality stays the single source of truth alongside
paths-ignore.

Net effect on the next scan: the non-shipped-code alerts auto-resolve, leaving
the security tab focused on shipped backend + frontend. No product code changes.

Deliberately NOT touched (assessed, left as-is):
- Stack-trace-exposure (detail=str(e) in routers) — these are intentional,
  helpful one-line diagnostics (the error-transparency work in b64f53b).
  Genericizing them on a loopback/single-user app would regress a product value
  for ~zero real benefit.
- "Critical" command-injection in exports.py and the high path-injections in
  settings.py / system.py — design-correct false positives: list-form argv (no
  shell), and the destination/model-dir/ffmpeg paths are arbitrary user-chosen
  paths by design (containment guards would break the features). Best handled by
  dismiss-with-justification, not code surgery.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-31 11:02:08 +05:30

190 lines
6.9 KiB
YAML

# Security scanning — runs on every PR, on push to main, and weekly.
#
# Complements the CodeRabbit + Greptile app reviews (which fire on PR creation)
# with deterministic, gating checks:
# • gitleaks — secret scanning. HARD FAIL: a leaked credential blocks merge.
# • CodeQL — Python + JS/TS SAST. Results land in the Security tab.
# • bandit — Python SAST (SARIF → Security tab). Reporting, non-gating.
# • pip-audit — Python dependency advisories. Reporting, non-gating.
# • bun audit — frontend dependency advisories. Reporting, non-gating.
#
# Only the secret scan gates the PR. Dependency advisories and bandit findings
# are surfaced as signal (Security tab / job log) rather than blocking every PR
# on a transitive upstream advisory — consistent with the "no ceremony,
# continuous-to-main" cadence.
name: Security
on:
pull_request:
branches: [main]
push:
branches: [main]
schedule:
# Mondays 06:00 UTC — catch advisories disclosed since the last PR.
- cron: "0 6 * * 1"
workflow_dispatch:
env:
# Match ci.yml: run JS actions on Node 24 (GH removes Node 20 in Sep 2026).
FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: true
# Least privilege by default; jobs that upload SARIF opt into security-events.
permissions:
contents: read
concurrency:
group: security-${{ github.ref }}
cancel-in-progress: true
jobs:
# ── Secret scanning (gating) ─────────────────────────────────────────────
# Full-history scan on push to main; PR-diff scan on pull_request (faster,
# and the action picks the right mode from the event automatically).
secrets:
name: Secret scan (gitleaks)
runs-on: ubuntu-22.04
steps:
- uses: actions/checkout@v4
with:
# gitleaks needs full history to scan all commits on push events.
fetch-depth: 0
# No authed git needed after clone; don't persist GITHUB_TOKEN.
persist-credentials: false
- name: gitleaks
uses: gitleaks/gitleaks-action@v2
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
# GITLEAKS_LICENSE is only required for GitHub *organizations*; this is
# a personal public repo, so the action runs free without it.
# ── CodeQL SAST (Python + JS/TS) ─────────────────────────────────────────
codeql:
name: CodeQL (${{ matrix.language }})
runs-on: ubuntu-22.04
permissions:
contents: read
security-events: write
strategy:
fail-fast: false
matrix:
language: [python, javascript-typescript]
steps:
- uses: actions/checkout@v4
with:
persist-credentials: false
- name: Initialize CodeQL
uses: github/codeql-action/init@v3
with:
languages: ${{ matrix.language }}
# Both targets are interpreted — no compiled build step needed.
build-mode: none
# Scope analysis to shipped product code. The excluded trees never
# ship in the installer's runtime path and produce the bulk of the
# note-level + false-positive findings (file-not-closed in eval
# harnesses, unused alembic migration globals, bind-all in tests,
# path sinks in the legacy Gradio research UI). Queries live in the
# inline config so there's a single source of truth next to
# paths-ignore. paths-ignore is supported here because build-mode is
# `none` (interpreted analysis).
config: |
queries:
- uses: security-and-quality
paths-ignore:
- omnivoice/eval
- research
- tests
- backend/migrations
- "**/*.test.js"
- "**/*.test.jsx"
- "**/*.test.ts"
- "**/*.test.tsx"
- name: Analyze
uses: github/codeql-action/analyze@v3
with:
category: "/language:${{ matrix.language }}"
# ── Python SAST (bandit → SARIF) ─────────────────────────────────────────
bandit:
name: Python SAST (bandit)
runs-on: ubuntu-22.04
permissions:
contents: read
security-events: write
steps:
- uses: actions/checkout@v4
with:
persist-credentials: false
- name: Setup Python 3.11
uses: actions/setup-python@v5
with:
python-version: "3.11"
# -ll: report MEDIUM+ severity only. -ii: MEDIUM+ confidence only.
# Keeps the SARIF focused on findings worth a human look. The scan step
# is allowed to "fail" (findings present) without failing the job; the
# SARIF upload still runs so results reach the Security tab.
#
# NOTE: the `sarif` output format lives in the `bandit[sarif]` extra
# (pulls in sarif-om + jschema-to-python). Plain `bandit` rejects
# `-f sarif`, so install via the extra spec.
- name: Run bandit
continue-on-error: true
run: |
pipx run --spec 'bandit[sarif]' bandit -r backend/ -ll -ii -f sarif -o bandit.sarif
# continue-on-error: this job is reporting-only. If bandit can't write a
# SARIF for any reason (no findings dir, pipx hiccup), don't fail the job.
- name: Upload bandit SARIF
uses: github/codeql-action/upload-sarif@v3
if: always()
continue-on-error: true
with:
sarif_file: bandit.sarif
category: bandit
# ── Dependency advisories (reporting) ────────────────────────────────────
dependencies:
name: Dependency audit
runs-on: ubuntu-22.04
steps:
- uses: actions/checkout@v4
with:
persist-credentials: false
- name: Install uv
uses: astral-sh/setup-uv@v3
with:
enable-cache: true
cache-dependency-glob: "uv.lock"
- name: Setup Python 3.11
uses: actions/setup-python@v5
with:
python-version: "3.11"
# Audit the resolved Python environment. Non-gating: a transitive
# advisory with no fix available should not wall off every PR.
- name: pip-audit (Python)
continue-on-error: true
run: |
uv sync
uv run --with pip-audit pip-audit
# Pin a floor: `bun audit` was added in bun 1.2.x, so guarantee it exists.
- name: Setup Bun
uses: oven-sh/setup-bun@v1
with:
bun-version: "1.2"
# `bun audit` reports advisories against the frontend lockfile. Non-gating
# for the same reason; also tolerant of older bun without the subcommand.
- name: bun audit (frontend)
continue-on-error: true
working-directory: frontend
run: |
bun install --frozen-lockfile
bun audit