WIKI
WIKI/PRODUCT DOCUMENTATION
vulnwatch.tech ↗

VULNWATCH WIKI · PRODUCT MODEL

Website vulnerability intelligence,
explained as a system.

This is the descriptive and technical wiki for vulnwatch.tech. It documents the assessment model behind VulnWatch: how an authorized target becomes observable signals, evidence-aware analysis, prioritized findings, and a remediation feedback loop. The product includes AI Analyst, an AI-assisted analysis layer that helps explain signals, correlate evidence, and prepare risk context for human review.

DOCUMENT Product architectureSCOPE Authorized website assessmentsREADING TIME 08 min
i

Scope of this wiki. The diagrams describe the public product and workflow model. They are intentionally not a disclosure of private source code, internal endpoints, deployment topology, customer data paths, or operational security controls.

02 · SYSTEM VIEW

Assessment architecture

A controlled, evidence-aware workflow from authorization to re-assessment. The layers are separated so that collection, interpretation, reporting, and follow-up remain understandable and reviewable.

VULNWATCH / ASSESSMENT SYSTEMPUBLIC PRODUCT MODEL
01CONTROL
PLANE
AUTHORIZATIONAsset owner / operatorlegitimate scope and intent
→
TARGET DEFINITIONDomain · URL · assetboundary and exclusions
→
PREVIEW GATEReview before executionapprove · adjust · stop
↓ approved assessment job
02COLLECTION
PLANE
ORCHESTRATIONControlled assessment jobbounded signal collection
HTTPresponses · redirects · methods
TLScertificates · protocols · posture
HEADERSbrowser controls · cookies
RECON / TECHpublic surface · technology clues
→
RAW SIGNALSObservable observationssource-specific evidence
↓ normalized observation set
03ANALYSIS
PLANE
NORMALIZATIONCommon finding shapededupe · classify · preserve source
→
EVIDENCEContext recordtimestamp · request · limitation
→
CORRELATIONSignal relationshipsconfidence · exposure context
→
AI ANALYSTAssisted interpretationexplain · correlate · contextualize
↓ contextualized finding candidates
04DECISION
PLANE
HUMAN REVIEWValidate meaningarchitecture · threat model · limits
→
RISK PRIORITYSeverity · impact · next actioncontext over raw volume
→
FINDING SETEvidence · rationalestructured for triage
↓ decision-ready security output
05OUTCOME
LOOP
REPORTWhat · why · where · nextactionable, readable output
→
REMEDIATIONInvestigate and fixconfiguration · code · process
→
RE-SCANVerify observable changerepeat within authorized scope

03 · SIGNAL PIPELINE

How observations become useful

VulnWatch is not a raw-output relay. Each stage adds a different kind of context, while preserving the distinction between an observable signal and a confirmed security conclusion. AI Analyst supports this process; it does not replace evidence or human judgment.

STAGEQUESTIONTECHNICAL DETAILRESULT
ScopeWhat may be assessed?Target, intent, boundaries, exclusions, preview checkpoint.Authorized job
ObserveWhat is externally visible?HTTP/TLS behavior, headers, cookies, public surface and technology signals.Raw observations
NormalizeWhat belongs together?Common finding shape, source preservation, classification and deduplication.Comparable records
AI AnalystWhat could the signals mean together?AI-assisted explanation, evidence correlation, confidence context and analyst-facing risk interpretation.Context for review
PrioritizeWhat deserves attention first?Severity, impact, exposure context, asset importance and practical next action.Triage order
VerifyDid the signal change?Remediation feedback and a new authorized assessment for comparison.Follow-up state

04 · OUTPUTS

Built for decisions,
not alert volume.

Reports are intended to make technical findings easier to review and act on. A useful result connects an observation to evidence, context, impact, and a practical next step.

01

Finding context

What was observed, where it was observed, and which signal layer produced it.

02

Risk rationale

Why the observation may matter, with confidence and limitations kept visible.

03

Action path

What to investigate, remediate, or verify next — without pretending to provide a universal fix.

05 · BOUNDARIES

Public wiki. Private core.

This documentation is the public descriptive layer for vulnwatch.tech. It is a window into the product model, not a copy of the production application.

PUBLIC WIKIWorkflow · principles · product model

This GitHub Pages site, README, safe diagrams and future sanitized integration documentation.

controlled
interface →
PRIVATE PRODUCT COREApplication · operations · customer data

Production implementation, deployment configuration, internal endpoints and operational controls remain protected.

Responsible use: VulnWatch is intended for authorized security testing only. Do not scan systems you do not own or have explicit permission to assess. Results are decision support, not a guarantee of complete security or coverage.

06 · ECOSYSTEM

Public tooling & interfaces

VulnWatch keeps its production core private, but publishes selected reusable pieces around the platform. Each element below is a public interface into the system — a repository to use, an action to run in CI, an MCP server to plug into AI agents, or live API documentation to integrate against.

WIKI · THIS SITEPublic overview

vulnwatch-project

The public project overview and the wiki you are reading: the assessment model, the signal pipeline, boundaries, and the documentation layer that describes how VulnWatch works without exposing the production application.

  • GitHub Pages documentation
  • Product & architecture model
github.com/anton-kulyk/vulnwatch-project ↗
OPEN SOURCE · SCANNINGDocker Compose

vulnwatch-proxy

A standalone authenticating HTTP proxy for scanner infrastructure, built on 3proxy. It boots on any Unix VPS with one command and exposes two listeners: one with Basic-auth for tools like curl, Nuclei and WPScan, and one without auth — ACL-locked to your scanner IP — for clients such as Chrome and Puppeteer that cannot send proxy credentials.

  • Stable, consistent exit IP per request
  • Segments scanner traffic from the local network
github.com/anton-kulyk/vulnwatch-proxy ↗
OPEN SOURCE · CIGitHub Action

vulnwatch-scan

A GitHub Action that runs a VulnWatch assessment straight from your pipeline. It submits an authorized target to the VulnWatch API with a token, waits for completion, reports finding counts by severity plus a risk score, can fail the job at a severity threshold you choose, and can save the raw report JSON as a build artifact.

  • anton-kulyk/vulnwatch-scan@v1
  • Fail on threshold (fail_on)
github.com/anton-kulyk/vulnwatch-scan ↗
OPEN SOURCE · MCPMCP server

mcp-server-vulnwatch

A Model Context Protocol server that brings VulnWatch into AI coding agents and assistants. It bridges the Developer API over stdio, so an agent can read account stats and scan history, submit an authorized target, track progress, and pull the finished report — using the same token model as every other integration.

  • vw_get_stats · vw_scan_status · vw_scan_report
  • Scan submission is opt-in (VULNWATCH_MCP_ALLOW_SCAN)
github.com/anton-kulyk/mcp-server-vulnwatch ↗
INTERFACE · LIVESwagger UI

Developer API docs

The interactive reference for the VulnWatch Developer API — authentication model, endpoints, request and response schemas, and the token flow that CI integrations such as vulnwatch-scan use to start scans and read results.

  • Token-authenticated
  • No guest path on scan endpoints
app.vulnwatch.tech/api/docs ↗