Web Pentest logo

Web Pentest

Community
hypnguyen1209
web-pentest

Use when pentesting a web application or API — injection, XSS/CSP, SSRF/cloud-metadata, HTTP desync & cache poisoning, SSTI/prototype-pollution/deserialization, JWT/OAuth/GraphQL/IDOR, business logic & single-packet race

Overview

Publisherhypnguyen1209
Repositoryoffensive-claude
Skill nameweb-pentest
Stars
396
Forks
71
Bundled files
18
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 18 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by hypnguyen1209 on GitHub. Read the source before you install it.

Installation

Install the Web Pentest AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/hypnguyen1209/offensive-claude/tree/main/skills/web-pentest
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/hypnguyen1209/offensive-claude.git /tmp/offensive-claude
mkdir -p .claude/skills
cp -r /tmp/offensive-claude/skills/web-pentest .claude/skills/web-pentest
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Web Pentest in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Web Pentest on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Web Pentest is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Web Application Penetration Testing

When to Activate

  • Web application security assessment (black/grey/white box)
  • API security testing (REST, GraphQL, WebSocket)
  • Authentication & session management testing (JWT, OAuth/OIDC, SAML)
  • Business-logic and race-condition hunting
  • WAF/CSP bypass, filter evasion, and CDN-layer attacks (desync, cache)
  • Validating modern 2024-2026 vectors: HTTP/1.1 desync (0.CL/TE.0), SSRF->cloud metadata, prototype-pollution->RCE, SSTI sandbox escapes

Input-Signal Routing (what to test when you see X)

Fast test selection — map an observed parameter/behavior to the right deep-dive:

Input signal you observeTest forGo to
id, uid, order_id, account, GUID/sequential refIDOR/BOLA, access controlreferences/auth-api-access-control.md
url, next, callback, dest, webhook, image_urlSSRF + cloud metadata, open redirect, OAuth chainreferences/ssrf-cloud-metadata.md
price, qty, coupon, balance, status, state changebusiness logic + single-packet racereferences/http-desync-cache.md (race)
template/preview/name reflected in outputSSTI -> RCE, reflected XSSreferences/ssti-deserialization.md
search/filter/sort reflected; q=, error echoes inputSQLi, XSS, command injectionreferences/injection-sqli-cmdi.md
token, jwt, Authorization, Cookie, OAuth stateJWT forgery, session/OAuth flawsreferences/auth-api-access-control.md
empty/missing/session=; login accepts no creds; user[$ne]=auth-check tampering (empty/null/NoSQL/type-juggle/HPP), token predictionreferences/auth-api-access-control.md
serialized blob / base64 object / __proto__deserialization, prototype pollutionreferences/ssti-deserialization.md
behind CDN/cache; X-Forwarded-*; keyed paramsdesync (0.CL/TE.0), cache poisoningreferences/http-desync-cache.md

Technique Map

TechniqueATT&CKCWEReferenceScript
SQL injection (UNION/blind/NoSQL)T1190CWE-89references/injection-sqli-cmdi.mdscripts/ssrf_toolkit.py (probe pattern) / sqlmap
OS command / argument injectionT1059CWE-78 / CWE-88references/injection-sqli-cmdi.md-
Reflected/stored/DOM XSST1059.007CWE-79references/xss-csp-clientside.mdscripts/xss_csp_forge.py
CSP bypass / script gadgetsT1059.007CWE-1021references/xss-csp-clientside.mdscripts/xss_csp_forge.py
Client-side prototype pollution -> XSST1059.007CWE-1321references/xss-csp-clientside.mdscripts/xss_csp_forge.py
SSRF + cloud metadata theftT1552.005CWE-918references/ssrf-cloud-metadata.mdscripts/ssrf_toolkit.py
IMDSv2 bypass (Axios gadget)T1552.005CWE-113references/ssrf-cloud-metadata.mdscripts/proto_ssti_gadgets.py
HTTP request smuggling / desyncT1190CWE-444references/http-desync-cache.mdscripts/desync_probe.py
Web cache poisoning / deceptionT1557CWE-525references/http-desync-cache.mdscripts/desync_probe.py
SSTI -> RCE (Jinja2/Twig/FM/...)T1059CWE-1336references/ssti-deserialization.mdscripts/proto_ssti_gadgets.py
Server-side prototype pollution -> RCET1059CWE-1321references/ssti-deserialization.mdscripts/proto_ssti_gadgets.py
Insecure deserialization (Java/.NET/PHP)T1059CWE-502references/ssti-deserialization.mdysoserial / ysoserial.net
JWT forgery (none/confusion/jku/jwk/kid)T1078CWE-347references/auth-api-access-control.mdscripts/jwt_forge.py
Auth-check tampering (empty/null/NoSQL-op/type-juggle/HPP)T1556CWE-287 / CWE-697 / CWE-943references/auth-api-access-control.md-
Session token prediction (weak PRNG/entropy)T1539CWE-330references/auth-api-access-control.mdscripts/session_entropy.py
OAuth/OIDC/SAML auth bypassT1078CWE-287references/auth-api-access-control.mdscripts/jwt_forge.py
GraphQL abuse (batch/alias/depth)T1190CWE-770references/auth-api-access-control.mdscripts/graphql_audit.py
IDOR / BOLA / mass assignmentT1078CWE-639 / CWE-915references/auth-api-access-control.md-
Race condition (single-packet)T1190CWE-362references/auth-api-access-control.mdscripts/race_single_packet.py
Business-logic abuse (price/refund/workflow/race)T1539CWE-841references/business-logic-abuse.mdscripts/race_single_packet.py
HTTP Parameter Pollution (HPP)T1190CWE-235 / CWE-88references/injection-sqli-cmdi.md-
File-upload -> webshell / RCET1505.003CWE-434references/injection-sqli-cmdi.md-
WAF bypass (cross-cutting evasion)T1190CWE-693references/waf-bypass.md-
XXE (file read / SSRF / OOB exfil / CI-CD)T1190CWE-611references/xxe.md-
CORS misconfigurationT1557CWE-942references/auth-api-access-control.md-
CSRF (cross-site request forgery)T1185CWE-352references/auth-api-access-control.md-

Quick Start

bash
# 0. Intake from recon-osint: hosts, tech fingerprint, endpoints, params.
#    Spider + content discovery
ffuf -u https://t/FUZZ -w raft-large.txt -mc all -fc 404 -o ffuf.json
nuclei -u https://t -severity critical,high -tags cve,exposure,misconfig

# 1. Injection — automated SQLi, manual cmdi
sqlmap -r request.txt --batch --level 3 --risk 2 --tamper=between,space2comment

# 2. XSS / CSP
python3 scripts/xss_csp_forge.py contexts --collector atk.tld
python3 scripts/xss_csp_forge.py csp "$(curl -sI https://t | grep -i content-security)"

# 3. SSRF -> cloud metadata
python3 scripts/ssrf_toolkit.py probe "https://t/api/fetch?url=FUZZ" --collab abc.oast.fun
python3 scripts/ssrf_toolkit.py cloud --provider aws

# 4. HTTP desync (2024-2026 vectors: cl0/expect/options/te0)
python3 scripts/desync_probe.py https://t --probe all --verbose

# 5. SSTI / prototype pollution / deserialization
python3 scripts/proto_ssti_gadgets.py detect
python3 scripts/proto_ssti_gadgets.py ssti --engine jinja2 --cmd "id"
python3 scripts/proto_ssti_gadgets.py proto --cmd "id"

# 6. Auth & API
python3 scripts/jwt_forge.py hs-confuse "$JWT" --pubkey pub.pem --claim role=admin
python3 scripts/graphql_audit.py introspect https://t/graphql

# 7. Logic / race
python3 scripts/race_single_packet.py https://t/redeem -d "code=X" -n 30 -H "Cookie: s=..."

# -> feed each confirmed issue into templates/exploit/findings/ with
#    severity, CWE, CVSS, PoC, evidence, ATT&CK ID, remediation.

OPSEC & Detection (summary)

TechniqueTelemetry / IOCDetection (Sigma/EDR)OPSEC note
SQLi / cmdiSQL keywords, SLEEP/WAITFOR, web-svc spawning sh/curlwebserver regex Sigma; EDR child-shell from www-datatime-based is slow+noisy; --random-agent --delay, in-band reads first
XSS / CSP / client-PP<script, onerror=, __proto__ in params; CSP report-uri hitswebserver regex; Trusted-Types violation alertspersistent __proto__ can break shared cached pages — avoid
SSRF / metadatahits to 169.254.169.254 / metadata hosts from app tierproxy Sigma; GuardDuty credential-exfilstop at proving token retrieval; rotate collaborator subdomains
HTTP desync / cacheCL+TE together, obs-fold, Expect on OPTIONS to CDNproxy Sigma; reject ambiguous framingpoisons co-tenant traffic — scoped only, never POST-smuggle prod
SSTI / PP / deser{{ }}/${ }, NODE_OPTIONS, rO0AB, JNDI egresswebserver/body Sigma; EDR java->LDAP/RMIid probe first; reverse shells trip EDR; revert polluted config
JWT / OAuth / GraphQLalg:none, unknown jku/kid, multi-op GraphQL bodyapp Sigma; failed-verify spikesforged-JWT attempts alert; cap GraphQL brute wordlist/rate
Race (single-packet)burst of near-simultaneous mutating requestsrate-by-endpoint Sigma in 1s windowkeep -n low; document state mutated for cleanup

Deep Dives

  • references/injection-sqli-cmdi.md — SQLi (UNION/blind/error/NoSQL), OS command & argument injection, HTTP Parameter Pollution (HPP), file-upload -> webshell/RCE, Lambda pool-poisoning, ORM CVEs, Neo4j Cypher injection, JSON operator probes, WebSocket delivery, sqlmap tamper workflow, modern WAF/Unicode evasion.
  • references/waf-bypass.md — CROSS-CUTTING WAF evasion: encoding/normalisation chains, syntax-level obfuscation per-class, protocol-level (H2C/HTTP/3/WebSocket), ML-WAF evasion (Cyrillic homoglyph/adversarial padding/feature-weight), SNI/domain fronting, TLS fingerprint, per-WAF payloads, CVE-2023-50969.
  • references/xxe.md — file read, blind/OOB exfil, error-based, SSRF via XXE (cloud metadata jar://, redirect chain), file-upload XXE (SVG/DOCX/XLSX/EPUB), SAML XXE, CI/CD pipeline XXE (K8s/Jenkins/GitLab/GitHub Actions), filter bypass, parser hardening versions.
  • references/xss-csp-clientside.md — XSS contexts, DOM sinks, CSP bypass + script gadgets, client-side prototype pollution -> XSS, mXSS/sanitizer bypass.
  • references/ssrf-cloud-metadata.md — SSRF parser confusion, gopher/redis, AWS/GCP/Azure metadata, IMDSv2 reality + Axios bypass, HTML-to-PDF switch, 2024-2026 CVEs (Azure OpenAI, Oracle EBS, Next.js, ColdFusion).
  • references/http-desync-cache.md — CL.TE/TE.CL, TE.0 (2024), 0.CL/Expect & CVE-2025-32094 (Akamai), TE.TE chunk-ext, cache poisoning/deception ("Gotta cache 'em all", ACM CCS 2024), pipelining-vs-desync caveat.
  • references/ssti-deserialization.md — per-engine SSTI RCE, sandbox escapes (CVE-2024-22195, Frappe/XWiki/Yeti), Node prototype-pollution -> RCE (GHunter, NODE_OPTIONS --import), Java/.NET/PHP deserialization gadget chains.
  • references/auth-api-access-control.md — JWT forgery, OAuth/OIDC/SAML, GraphQL batch/alias/depth abuse, session & auth-check tampering (empty/null/falsy value, NoSQL-operator + type-juggling + HPP into auth), session token prediction/entropy (scripts/session_entropy.py), test-account setup for cross-identity testing, IDOR/BOLA/mass assignment, CORS misconfiguration, CSRF, and the HTTP/2 single- packet race attack.
  • references/business-logic-abuse.md — model the intended state machine, then break transitions: price/quantity/coupon tampering, refund/payout & single-packet races, workflow-step skip, idempotency gaps, tenant/role confusion, flaw-chaining.
  • references/proxy-mcp-integration.md — driving a live Burp/Caido proxy MCP for ground-truth traffic + replay-with-auth, with mandatory header redaction (redact_headers.py) at the data boundary.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Web Pentest AI skill do?

Use when pentesting a web application or API — injection, XSS/CSP, SSRF/cloud-metadata, HTTP desync & cache poisoning, SSTI/prototype-pollution/deserialization, JWT/OAuth/GraphQL/IDOR, business logic & single-packet race

Why use Web Pentest on TypingMind?

Because you install it once and use it with any model. Web Pentest is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Web Pentest in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/hypnguyen1209/offensive-claude/tree/main/skills/web-pentest. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Web Pentest?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Web Pentest?

As many as you like. As long as a model supports skills, you can use Web Pentest with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Web Pentest AI skill free?

Yes. It is published on GitHub by hypnguyen1209 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇