pablopunk Security Audit
Use this skill for defensive security work on Nevermind. Produce evidence-backed findings, not generic advice.
First steps
- Establish scope from the repo. Identify the desktop app, backend, deployed surfaces, auth providers, data stores, IPC boundaries, extension APIs, and privileged OS actions before judging risk.
- Check local guidance. Read
AGENTS.md, then the relevant project skill files. For Electron-specific work, read../electron-best-practices/SKILL.mdand linked files. - Map trust boundaries. Renderer ↔ preload ↔ main, extensions ↔ host API, user content ↔ extension views, desktop ↔ backend, backend ↔ upstream AI providers, auth sessions ↔ API tokens, admin ↔ user.
- Audit high-impact paths first. Auth, tokens, admin APIs, proxy/billing, custom protocols, shell/file/URL actions, IPC handlers, extension permissions, secrets/config, update/release pipeline.
- Verify findings. Prefer code references, concrete reproduction steps, safe local tests, and exact commands over speculation.
Nevermind security surfaces
- Electron main/preload/renderer:
src/app/electron/main.ts,src/app/electron/preload.ts,src/app/electron/palette-window.ts,src/app/palette/preload-api.ts. - Extension host and API:
src/app/resources/nevermind-extension-api.d.ts,src/app/electron/main.ts,src/app/palette/extension-view.tsx, internal extensions undersrc/app/resources/. - Desktop auth/model proxy integration:
src/app/electron/nevermind-auth.ts,src/app/electron/ai.ts. - Astro backend routes:
src/backend/src/pages/api/**, especially auth, device auth, admin, tokens, and/api/v1/**proxy routes. - Backend libraries:
src/backend/src/lib/workos.ts,tokens.ts,admin.ts,proxy.ts,ratelimit.ts,cron-auth.ts,audit.ts,settings.ts. - Persistence/config:
src/backend/src/db/schema.ts, Drizzle migrations,electron-builder.yml,src/backend/vercel.json, env examples.
Audit checklist
- Auth/session: redirect handling, sealed cookie settings, CSRF on cookie-authenticated mutations, logout, device-code approval/exchange, role escalation.
- API tokens: generation entropy, prefix leakage, hashing, revocation, last-use writes, bearer extraction, rate limits, token exposure in logs/UI.
- Authorization: admin gates, object ownership checks, extension permission checks, IPC caller assumptions, path/URL action ownership.
- Electron hardening: context isolation, sandbox, node integration, navigation/window-open policy, permission handlers, CSP/custom protocol behavior, preload API input validation.
- Extension isolation: least-privilege permissions, shell/system actions, file and app helpers, iframe sandbox, untrusted HTML/media, clipboard and AI permissions.
- Backend proxy/billing: transparent auth swap, forwarded headers, request size/token limits, upstream URL construction, streaming parsing, abuse/rate limits, billing failure modes.
- Input/output safety: SSR/script injection, DOM insertion, markdown/rendered content, URL/path validation, shell argument handling, log redaction.
- Secrets/config: committed env files, Vercel env exposure, Sentry DSNs, WorkOS/Redis/Neon keys, release signing material, local auth storage.
- Supply chain: dependency vulnerabilities, Electron version posture, packaged resources, updater/release artifact trust, native modules.
Report format
For a full report, include:
- Scope and methodology.
- Architecture/trust-boundary summary.
- Executive summary with severity counts.
- Findings sorted by severity, each with: title, severity, affected files/routes, evidence, impact, reproduction or exploit sketch, recommended fix, verification command/test.
- Positive controls already present.
- Residual risks and follow-up hardening roadmap.
Do not print secrets, live tokens, cookies, raw user data, or production payloads. Redact values while preserving enough shape to prove the issue.