JAVASCRIPT / TYPESCRIPT REPOSITORY ANALYSIS

Review AI-built JavaScript and TypeScript projects before release

Trace important production flows, observed security controls, and engineering risks without executing project code or asking an AI model to decide the findings. The core review comes from deterministic analysis and versioned engineering knowledge.

We inspect the declared supported repository scope, connect functions and production flows, evaluate engineering controls, and retain traceable evidence. This is not a sample of files or an AI guess.

See 4 real repository results

Private source handling: we use your repository only for this analysis. The archive is deleted after extraction and the source workspace is deleted when the run ends. We retain only the generated results and minimal run metadata for your secret link for 24 hours; download and store your Customer Report and Prompt Pack within that time. Source code is not used for training or secondary use.

REAL REPOSITORIES · PINNED RESULTS

See what gets checked before an AI-built app ships

Four public repositories, four published evidence-backed results you can open. We pin the source, reconstruct the important paths static analysis can support, separate confirmed issues from follow-up work, and validate every published artifact.

01Pin the sourceAn immutable public commit keeps the result reproducible.
02Trace behaviorRoutes, functions, controls, effects, data, and provider boundaries.
03Bound the claimsFacts, recommendations, analyzer gaps, and unknowns stay distinct.
04Validate deliveryReport, dossier, actions, and Prompt Pack must reconcile.

01Public JS/TS reference

Report publishedPinned 513da90Quality checks passed

Next.js 16 full-stack template

A compact full-stack starter used to test whether FlowPreflight stays cautious when a repository contains framework capability but limited product-specific behavior.

Next.js 16TypeScriptNextAuthMongoDB

1important production flows

10follow-up checks

1evidence boundaries

54 files inventoried · 661 source facts connected · 30 functions analyzed

Why this result matters

Shows why even a clean starter needs a boundary review before real product logic is layered on top.

See what we found Read actual report Source repository

02Public JS/TS reference

Report publishedPinned f0a5c55Quality checks passed

Play Next.js SaaS starter

A medium SaaS codebase with identity, email/OTP, content, database, and Stripe boundaries—useful for showing how the report separates several product journeys.

Next.jsTypeScriptNextAuthPrisma

4important production flows

13follow-up checks

1evidence boundaries

157 files inventoried · 3,031 source facts connected · 128 functions analyzed

Why this result matters

Demonstrates the value of tracing auth and payment boundaries before a SaaS starter becomes a production service.

See what we found Read actual report Source repository

03Built with Lovable

Report publishedPinned 049db25Quality checks passed

Lovable-generated Gantt planner

A larger AI-generated project planner with persistent CRUD, project membership, roles, task ordering, and a prompt history that documents how the app was assembled.

ReactTypeScriptViteSupabase

Openflow reconstruction boundary

3follow-up checks

1evidence boundaries

135 files inventoried · 3,538 source facts connected · 140 functions analyzed

Why this result matters

Shows the exact kind of integration boundary an AI builder can miss when a polished UI is ready before its deploy assumptions are proven.

See what we found Read actual report Source repository

04Initial UI created with v0

Report publishedPinned 502ed5bQuality checks passed

v0-assisted Anthropic token counter

A small AI-assisted utility that sends user text through a server route to Anthropic’s token-counting API—ideal for testing provider, input, logging, and secret-reference boundaries.

Next.js 14TypeScriptAnthropic SDKReact

2important production flows

1follow-up checks

1evidence boundaries

24 files inventoried · 637 source facts connected · 17 functions analyzed

Why this result matters

A small app can still cross a sensitive AI-provider boundary; this demo shows why “the feature works” is not the same as “the deploy assumptions were reviewed.”

See what we found Read actual report Source repository

YOUR REPOSITORY IS NEXT

AI helped you build it. Let evidence review it before deploy.

Upload a focused ZIP or analyze a public GitHub repository. Source is never executed.

Analyze your repository

RETURN TO AN ANALYSIS

Open a saved result

Paste the secret link from a previous analysis. The link includes its access key and works for 24 hours after completion.

UPLOAD PREPARATION

Prepare a focused analysis ZIP

Run this from the repository root to exclude dependencies, generated output, caches, test files by default, media, and known secret-file patterns before upload.

Requires PowerShell 7 and verifies the downloaded packer with SHA-256 before execution.

Loading script…

When it finishes, upload project.flowpreflight.zip above.

A REPORT YOU CAN ACT ON

Start with the decision. Open the evidence when you need it.

The new report system turns a detailed repository scan into a readable product review without throwing the underlying evidence away. Founders and product owners can understand the result first; engineers can trace every important statement back to flows, functions, routes, controls, and limitations.

01
Understand the product
See what the repository appears to build, which capabilities were detected, and which customer journeys matter most.
02
Read the overall result
Separate confirmed issues from follow-up checks. See what looks good without treating missing runtime evidence as a source-code defect.
03
Inspect important flows
Follow routes, functions, data operations, external effects, and observed controls through customer-friendly diagrams and flow summaries.
04
Plan the next move
Use prioritized recommendations and scoped Prompt Pack tasks with implementation and independent verification instructions.

Observed evidence

A source-backed fact or control the analyzers could trace.

Follow-up check

A focused question worth reviewing—not automatically a defect.

Evidence boundary

A path or runtime question the available static evidence cannot close.

Confirmed issue

A condition supported by closed, traceable evidence and shown separately.

YOUR REPORT BUNDLE

Four views of the same validated analysis

Each view is assembled from the same canonical evidence. Counts, priorities, limitations, and references must reconcile before publication; a validation note stays attached instead of silently removing report content.

01 / CUSTOMER REPORT

The product and release reading view

Explains the product, important journeys, observed controls, confirmed issues, follow-up checks, architecture signals, and recommended engineering phases in plain English.

Best for

Founders, product owners, and engineering leads
Understanding what was actually inspected
Separating evidence from recommendations and unknowns
Choosing which review area to open next

02 / VISUAL DIAGRAM PACK

See how the important journeys connect

Turns source-backed journeys and control paths into browser-readable diagrams, with explicit boundary notes when a complete path cannot be established.

Best for

Entrypoint-to-result journey maps
Control and side-effect placement
Application and integration relationships
Sharing system context with the team

03 / TECHNICAL FLOW DOSSIER

Trace the report back to source evidence

Provides the flow-by-flow technical projection behind the customer reading view without exposing internal analyzer databases.

Best for

Critical Function Projection and flow index
Resolved routes, files, functions, and calls
Controls, effects, and ordering evidence
Detected, partial, and coverage-bound outcomes

04 / PROMPT PACK

Hand scoped work to an AI coding tool

Converts only evidence-backed actions into bounded implementation and independent verification tasks, while recommendations and unknowns remain labeled as such.

Best for

Shared Agent Policy and prioritized plan index
Exact scope and evidence context
Implementation checkpoints
Separate IMPLEMENT and VERIFY instructions

DETERMINISTIC REVIEW · VERSIONED KNOWLEDGE

Not an AI opinion about a sample of your code

FlowPreflight’s core review does not ask an AI model to inspect the repository or decide what is wrong. Purpose-built analyzers examine the complete supported static scope, reconstruct every flow they detect, and evaluate that evidence against versioned engineering knowledge.

01
Inventory every in-scope file
Enumerate supported JS/TS source plus relevant manifests and configuration. Exclusions, limits, and parse failures remain explicit.
02
Resolve how the code connects
Link imports, symbols, calls, routes, inputs, controls, data operations, provider calls, and side effects where static resolution succeeds.
03
Reconstruct every detected flow
Trace entrypoint-to-result paths. When a path cannot be closed, retain the available evidence and the exact analysis boundary instead of guessing.
04
Apply versioned knowledge
Compare each flow with system- and flow-specific engineering checklists for identity, payments, webhooks, data writes, queues, AI providers, and other detected capabilities.
05
Validate every published claim
Reconcile counts and references, distinguish facts from recommendations and unknowns, and run privacy, contradiction, and artifact-quality gates.

ASKING AN AI MODEL

Context-limited narrative review

Usually reads selected files or chunks that fit the supplied context.
Can overlook files, indirect call paths, repeated patterns, or relationships outside that context.
May produce a persuasive explanation without a reproducible source-to-effect evidence path.
Useful for exploration and suggestions, but the answer can vary between prompts and runs.

FLOWPREFLIGHT CORE ENGINE

Repeatable evidence review

Processes every file admitted to the declared supported static scope instead of choosing a conversational sample.
Preserves every detected flow, resolved relationship, limitation, and unresolved boundary in structured artifacts.
Uses compiled knowledge packs to determine which controls and engineering questions apply to each flow type.
Produces traceable findings and follow-up checks whose evidence references must validate before publication.