recon

name	recon
description	Systematically discover alternative endpoints, bypasses, and undocumented access paths on a target platform. Use after the normal flow works to find simpler or more efficient approaches.
allowed-tools	mcp__chrome-devtools__*, mcp__powhttp__*

Recon

Systematically discover alternative endpoints, bypasses, and undocumented access paths. Use this after you have a working flow (via browser-navigate + network-debug) and want to find simpler or more efficient approaches.

1. Map the Attack Surface

Subdomain enumeration

• Certificate Transparency — https://crt.sh/?q=%25.example.com&output=json lists all issued certificates
• Reverse DNS — dig -x <IP> to find co-hosted domains on the same server
• DNS records — dig example.com ANY for MX, TXT, CNAME revealing internal hosts
• Response headers — X-Served-By, X-Backend-Server, Via in captured traffic
• JavaScript bundles — grep captured JS for hardcoded URLs and subdomains
• CORS headers — Access-Control-Allow-Origin reveals sibling domains

Path enumeration

• robots.txt — disallowed paths often point to interesting endpoints
• sitemap.xml — URL structure and available pages
• OpenAPI/Swagger — /swagger.json, /openapi.json, /api-docs, /v1/docs
• GraphQL — /graphql, /graphiql. Run introspection query if found
• Health/status — /health, /status, /version, /info

Tech stack identification

• Response headers — X-Powered-By, Server, X-Cache
• Cookie names — JSESSIONID (Java), PHPSESSID (PHP), connect.sid (Express)
• Error pages — trigger 404/500 to identify framework
• HTML source — __NEXT_DATA__ (Next.js), __NUXT__ (Nuxt), generator meta tags

2. Alternative Access Surfaces

• Mobile APIs — proxy a phone, compare mobile vs desktop responses
• Mobile web — m.example.com or responsive variants
• Export/bulk endpoints — /export, /download, ?format=csv
• Print views — /print, ?print=true
• RSS/Atom — /feed, /rss, /atom.xml
• Developer/partner APIs — developer.example.com, /developers

3. API Version and Format Exploration

• Version probing — try v1, v2, v3; check for version headers (API-Version, Accept-Version)
• Format negotiation — Accept: application/json on HTML pages, JSONP via ?callback=x
• GraphQL alongside REST — same data, different access control or rate limits
• Content-Type alternatives — form-encoded vs JSON POST bodies may have different validation

4. Embedded Data Extraction

Pages often embed all their data in HTML, avoiding the need for separate API calls.

In-browser extraction

// JSON data blocks
evaluate_script(function: "() => { return [...document.querySelectorAll('script[type=\"application/json\"]')].map(s => s.textContent.substring(0, 200)) }")

// Linked data
evaluate_script(function: "() => { return [...document.querySelectorAll('script[type=\"application/ld+json\"]')].map(s => JSON.parse(s.textContent)) }")

// Inline JSON assignments
evaluate_script(function: "() => { return [...document.querySelectorAll('script:not([src])')].map(s => s.textContent).filter(t => t.includes('window.__') || t.includes('initialState') || t.includes('pageData')).map(t => t.substring(0, 300)) }")

From captured traffic

• Script tags — <script type="application/json"> and <script type="application/ld+json"> in HTML responses. Use powhttp_query_body with JQ expressions to extract.
• Inline assignments — window.__DATA__ = {, "sectionPopupData":, var initialState = in script blocks
• Linked data URLs — scripts may reference internal API URLs not called during page load

Use powhttp_get_entry(body_mode: "full") on HTML page loads to inspect raw source.

5. JavaScript Source Analysis

• Bundle grep — search for baseURL, apiUrl, fetch(", axios.create in JS files
• Source maps — append .map to JS URLs to get original source
• Build-time env vars — NEXT_PUBLIC_, REACT_APP_, VUE_APP_ prefixed variables
• Webpack chunks — named chunks correspond to features, revealing endpoint patterns

6. Platform-Specific Endpoints

• Next.js — /_next/data/<buildId>/<path>.json returns page data as JSON
• Nuxt.js — /_payload.json for page data
• Shopify — /products.json, /collections.json, Storefront API
• WordPress — /wp-json/wp/v2/ REST API
• Elasticsearch — /_search, /_mapping if exposed
• Firebase — <project>.firebaseio.com/.json if rules allow

7. Related Site Analysis

• Same parent company — shared backends, consistent endpoint patterns
• Same platform — one site's patterns apply to all sites on that platform
• Community resources — GitHub for unofficial SDKs: site:github.com "example.com" api
• Wayback Machine — old JS bundles reveal historical endpoints that may still be active
• CORS headers — Access-Control-Allow-Origin reveals sibling domains

8. Auth and Session Shortcuts

Session inspection

// Get all cookies
evaluate_script(function: "() => { return document.cookie }")

// Inspect localStorage keys
evaluate_script(function: "() => { return Object.keys(localStorage) }")

// Get specific localStorage value
evaluate_script(function: "() => { return localStorage.getItem('authToken') }")

// Decode a JWT from a cookie
evaluate_script(function: "() => { const token = document.cookie.split('; ').find(c => c.startsWith('jwt=')); if (!token) return null; const payload = token.split('.')[1]; return JSON.parse(atob(payload.split('=')[0].replace(/-/g, '+').replace(/_/g, '/'))) }")

Session lifecycle

• Minimal cookie set — test subsets to find minimum required cookies. Use powhttp_get_entry(include_headers: true) to check which cookies the server actually reads.
• Token reuse — web auth tokens on mobile endpoints and vice versa
• JWT inspection — decode expiry (exp field), understand refresh lifecycle
• Silent refresh — GET homepage may renew session without re-auth
• Cookie dependency mapping — which cookies the server reads vs ignores
• Unauthorized detection — some sites signal expiry by clearing cookies (Set-Cookie with past expiration) rather than returning error status

Endpoint discovery from known patterns

• URL pattern inference — if /listings/getListings exists, try /sales/getSales, /orders/getOrders
• Query parameter exploration — test pageSize, limit, offset, sort, dateFrom, dateTo, status, filter
• Larger page sizes — default pagination is often conservative (20-50). Test pageSize=500 or pageSize=1000.

Response schema analysis

• Abbreviated field names — minified APIs use short keys ("i" for ID, "n" for name). Use powhttp_query_body to extract samples and infer meaning.
• Nullable fields — compare multiple responses with powhttp_describe_endpoint to see field consistency
• Enum collection — powhttp_query_body with deduplicate: true across a cluster to collect all status/type values
• Nested pagination — check top-level structure for wrapper objects ({"items": [...], "total": 500, "page": 1})

Keys in source

• apiKey, token, client_id in JS bundles or HTML
• Geo/locale headers — X-Forwarded-For, Accept-Language effects on responses