security-review · git:20260827.303fa2e · 2026-08-27 · sha256 314bf5ab3ab760d8
security-review git:20260827.303fa2eA
Immutable. This exact content never changes and is served at /api/v1/blob/314bf5ab3ab760d8.
No tool to install first
claude plugin marketplace add https://markdownregistry.com/mp/a/art_3meswfiz6wsjs5by/314bf5ab3ab760d875e1516ed6682c856df4d07e906b5c395b5ba908ee5f5c98.json
Then /plugin install security-review@markdownregistry-3meswfiz6wsjs5by in Claude Code. It installs this exact version, the one graded A here, pinned to the commit it was graded at, and leaves it there.
--- name: security-review description: Review Langfuse changes for SSRF, tenant isolation, secret handling, unsafe redirects or uploads, RBAC drift, and client telemetry privacy. Use when a design or change accepts URLs or host fields, handles secrets or cross-tenant data, makes outbound requests, adds an integration, follows redirects, widens permissions, or can send customer-controlled UI data through analytics, monitoring, or session replay. --- # Security Review Use this skill when reviewing or planning code that touches a security-sensitive surface in Langfuse. It collects the recurring findings the team has seen in external security reports so that future agents catch them at design and review time rather than after the fact. ## When to Apply Apply this skill when the change touches any of: - a user-supplied URL, host, endpoint, `baseURL`, or webhook target - a new outbound HTTP request (`fetch`, `axios`, AWS SDK client init with a custom `endpoint`, OpenAI/Anthropic/Bedrock client init with a custom `baseURL`, etc.) - a new integration form under Settings -> Integrations or any admin-configurable network destination - a new tRPC procedure or public API route that mutates project-scoped data or changes who can access it - secrets, API keys, signing secrets, or encryption-at-rest fields - redirect-following or cross-origin header handling - file uploads, image proxies, or other binary data flowing in or out - product analytics, browser monitoring, session replay, or another client-side telemetry path that can transmit customer-controlled content Apply this skill during **plan mode** when designing a new integration so the correct validation surfaces land in the plan, not in a follow-up CVE. ## How to Read This Skill 1. Open [references/checklist.md](references/checklist.md) and run the mental sweep against the change. 2. For each bullet that fires, open the matching topic reference. | Topic | Open when | File | | ----------------------------------- | ------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------- | | SSRF and outbound URL validation | The change accepts or fetches a user-supplied URL, host, or endpoint | [references/outbound-url-validation.md](references/outbound-url-validation.md) | | Secret read paths | The change shapes what a read route returns for an entity or config blob that also stores a credential | [references/secret-read-paths.md](references/secret-read-paths.md) | | Client telemetry and session replay | The change records DOM, browser state, events, logs, or network data in a third-party system | [references/client-telemetry-privacy.md](references/client-telemetry-privacy.md) | The catalog is intentionally short today. New topic files are added as new finding classes recur (see "Extending This Skill"). ## Output Expectations (Review Mode) When this skill is used during code review: - List findings first, ordered by severity, with file and line references. - For each finding, name the canonical helper or known-good call site the author should copy. - For SSRF-class findings, point at [references/outbound-url-validation.md](references/outbound-url-validation.md) rather than re-deriving the fix. - Call out missing **negative tests** (private-IP, cross-tenant, missing-scope) as findings, not as nice-to-haves. ## Output Expectations (Design / Plan Mode) When this skill is used while planning: - Restate which surfaces the new feature exposes (forms, public API routes, worker entrypoints). - For each surface that matches a checklist trigger, name the validator or helper that must be invoked and at which layer (save-time, use-time, connection-time, redirect-time). - Treat "we will validate later" as a design defect: validation belongs in the same change that introduces the surface. ## Extending This Skill Add a new `references/<topic>.md` whenever a security finding recurs across features or PR reviews. Keep each reference narrow and concrete: 1. Threat in plain language (one paragraph). 2. Canonical helpers in this repo, with paths. 3. Known-good call sites that can be copied. 4. Required defenses (save-time, use-time, transport-time, etc.). 5. Anti-patterns to flag in review. Then add a one-line trigger to [references/checklist.md](references/checklist.md) pointing at the new topic file, and add a row to the table above. Candidates for future references (do not add until a real finding recurs): - Tenant isolation (`projectId` filters across Prisma and ClickHouse) - Redirect mishandling and sensitive-header propagation - File upload validation and content-type sniffing - RBAC scope drift on new tRPC/public API endpoints - Signed URL scoping (expiry, path, method) - Public API rate limiting and auth boundary checks ## Integration With Other Skills - The shared `code-review` skill should defer here for any change that matches the triggers above; see [code-review/SKILL.md](../code-review/SKILL.md). - The shared `backend-dev-guidelines` skill should defer here when adding outbound HTTP, integration config, or URL-accepting procedures; see [backend-dev-guidelines/SKILL.md](../backend-dev-guidelines/SKILL.md). - Confirmed issues with reproduction evidence go through `linear-bug-triage` for Linear handoff.