ctf-web

ctf-web is a skill for Claude Code from ljagiello/ctf-skills. It costs 124 tokens per session (4,324 once invoked), scanned B, original, MIT.

A routing and reference guide for web-focused capture-the-flag security puzzles. It covers websites, APIs, browser code, login systems, file uploads, templates, and smart-contract interfaces.

In plain words
What is it for?
Mapping web targets and testing issues such as cross-site scripting, SQL injection, server-side request forgery, XML attacks, token flaws, authentication bypasses, and request smuggling.
Why use it?
It helps identify the application's trust boundaries and choose an appropriate way to investigate common web vulnerabilities.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: mentions Claude Code.

Needs its repository: it runs a file that does not travel with it, so clone the repository first. The line is bash scripts/install_ctf_tools.sh pat # PAT only.

Good fit Mapping web targets and testing issues such as cross-site scripting, SQL injection, server-side request forgery, XML attacks, token flaws, authentication bypasses, and request smuggling.

Compare 6 skills from other repositories ↓
About the project

ctf-skills is a collection of agent instructions for solving capture-the-flag security challenges, including web exploitation, binary vulnerabilities, cryptography, reverse engineering, forensics, and OSINT. It is intended for agents and people working through CTF problems, and its skills can be loaded into supported coding-agent workflows.

ljagiello/ctf-skills · 3,549 stars · on GitHub · ljagiello.github.io

Install

Getting it into your agent

It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.

Clone the repo
git clone --depth 1 https://github.com/ljagiello/ctf-skills
agentmods
npx agentmods add skills/ljagiello/ctf-skills/ctf-web

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for ctf-web

README.md
[![agentmods](https://agentmods.dev/badge/skills/ljagiello/ctf-skills/ctf-web/github.svg)](https://agentmods.dev/skills/ljagiello/ctf-skills/ctf-web)
Your own site
<a href="https://agentmods.dev/skills/ljagiello/ctf-skills/ctf-web"><img src="https://agentmods.dev/badge/skills/ljagiello/ctf-skills/ctf-web/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for ctf-web

Your own site · 80×15
<a href="https://agentmods.dev/skills/ljagiello/ctf-skills/ctf-web"><img src="https://agentmods.dev/badge/skills/ljagiello/ctf-skills/ctf-web.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 124 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,324 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 2 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • Socket fail 27 Oct 2026
  • Snyk pass 27 Oct 2026
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 3 findings, up to high

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • high YARA Match · line 88
    YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).
    Fix: Remove offensive tool references and exploit code. Legitimate agent skills should not contain penetration testing tools, exploit frameworks, or reconnaissance utilities.
  • medium Excessive Agency · line 90
    Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
    Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
  • medium Data Exfiltration · line 98
    Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
    Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00124 $0.04324
Opus 5.5 $0.00050 $0.01730
Sonnet 5.5 $0.00025 $0.00865
Haiku 4.5 $0.00012 $0.00432

Measured yesterday against content hash c521b3a35bc9, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade B, and why

ctf-web scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

The scan reads SKILL.md. This mod also ships 1 executable file (scripts/async_fuzz.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Sends data to an external URLmediumData exfiltration

A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.

curl -v -X POST https://target.com/api -H "Content-Type: application/json" -d '{}'

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

apt install hashcat jq curl
Origin

Copies of this mod

2 near-identical copies found in the catalogue:

ctf-web/SKILL.md · 246 lines

How it starts

The opening of the file, as written. The whole thing — 246 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CTF Web Exploitation

Use this skill as a routing and execution guide for web-heavy challenges. Keep the first pass short: map the app, confirm the trust boundary, and only then dive into the detailed technique notes.

Prerequisites

Python packages (all platforms):

pip install sqlmap flask-unsign requests httpx

Linux (apt):

apt install hashcat jq curl

macOS (Homebrew):

brew install hashcat jq curl

Go tools (all platforms, requires Go):

go install github.com/ffuf/ffuf/v2@latest

Manual install:

  • ysoserial — GitHub, requires Java (Java deserialization payloads)
  • PayloadsAllTheThings — git clone to ctf-web/payloads/PayloadsAllTheThings (auto via install script or lazy clone)
    bash scripts/install_ctf_tools.sh pat   # PAT only
    bash scripts/install_ctf_tools.sh all   # all tools including PAT
    # manual fallback:
    git clone --depth 1 https://github.com/swisskyrepo/PayloadsAllTheThings.git ctf-web/payloads/PayloadsAllTheThings
    

    PAT is optional and on-demand — not required at load time. The skill works without it (graceful degrade): pat-reference.md provides an offline index with exemplar payloads; bulk wordlists require the clone above.

MCP servers (optional): Playwright MCP (microsoft/playwright-mcp) for browser-dependent challenges (DOM XSS, admin bot, SPA client secrets) and Burp Suite MCP (PortSwigger/mcp-server) for wire-level work (header-level HTTP/1.1+HTTP/2 control, smuggling, blind SSRF/XXE via Collaborator, proxy-history regex mining) — see tools-mcp.md. Two things to do before relying on Burp MCP: start Burp before the agent, and turn off the two approval gates in Burp's MCP tab or the first tool call blocks on a modal dialog.

Additional Resources

  • sql-injection.md - SQL injection techniques: auth bypass, UNION extraction, filter bypasses, second-order SQLi, truncation, race-assisted leaks, INSERT ON DUPLICATE KEY UPDATE password overwrite, innodb_table_stats WAF bypass
  • server-side.md - PHP type juggling, php://filter LFI, Python str.format traversal, SSTI (Jinja2, Twig, ERB, Mako, EJS, Vue.js, Smarty), SSRF (Host header, DNS rebinding, curl redirect, unescaped-dot regex, SNI FTP smuggling, mod_vhost_alias), PHP hash_hmac NULL
  • server-side-2.md - XXE (basic, OOB, DOCX upload), XML injection via X-Forwarded-For, PHP variable variables, PHP uniqid predictable filename, sequential regex replacement bypass, command injection (newline, blocklist, sendmail CGI, multi-barcode, git CLI), GraphQL injection (introspection, batching, interpolation)
  • server-side-exec.md - Direct code execution paths, upload-to-RCE, deserialization-adjacent execution, LaTeX injection, header and API abuses
  • server-side-exec-2.md - More execution chains: SQLi fragmentation, path parser tricks, polyglot uploads, wrapper abuse, filename injection, BMP pixel webshell with filename truncation
  • server-side-deser.md - Java/Python/PHP deserialization and race-condition playbooks, PHP SoapClient CRLF SSRF via deserialization
  • server-side-advanced.md - Advanced SSRF, traversal, archive, parser, framework, and modern app-server issues, Nginx alias traversal
  • server-side-advanced-2.md - Docker API SSRF, Castor/XML, Apache expression reads, parser discrepancies, Windows path tricks, rogue MySQL server file read
  • server-side-advanced-3.md - Part 3 (CSAW/35C3/ASIS/PlaidCTF 2018): WAV polyglot upload, multi-slash URL path.startswith bypass, Xalan XSLT math:random() seed guess, SoapClient _user_agent CRLF method smuggling, gopher:/// no-host URL scheme bypass, SSRF credential leak via attacker-specified outbound URL
  • server-side-advanced-4.md - Part 4: WeasyPrint SSRF/file read (CVE-2024-28184), MongoDB regex/$where blind oracle, Pongo2 Go template injection, ZIP PHP webshell, basename() bypass, wget CRLF SSRF→SMTP, Gopher SSRF to MySQL blind SQLi, React Server Components Flight RCE (CVE-2025-55182), AMQP/TLS interception via sslsplit+arpspoof, CairoSVG XXE, Bazaar repo reconstruction
  • client-side.md - XSS, CSRF, cache poisoning, DOM tricks, admin bot abuse, request smuggling, paywall bypass
  • client-side-advanced.md - CSP bypasses, Unicode tricks, XSSI, CSS exfiltration, browser normalization quirks, postMessage null origin bypass
  • auth-and-access.md - Auth/authz bypasses, hidden endpoints, IDOR, redirect chains, subdomain takeover, AI chatbot jailbreaks
  • auth-and-access-2.md - Part 2 (2018-era): std::unordered_set bucket collision auth bypass, nodeprep.prepare Unicode homograph username collision, SRP A=0/A=N auth bypass, ArangoDB AQL MERGE privilege escalation
  • auth-jwt.md - JWT/JWE manipulation, weak secrets, header injection, key confusion, replay
  • auth-infra.md - OAuth/OIDC, SAML, CORS, CI/CD secrets, IdP abuse, login poisoning
  • node-and-prototype.md - Prototype pollution, JS sandbox escape, Node.js attack chains
  • web3.md - Solidity and Web3 challenge notes
  • cves.md - CVE-driven techniques you can match against challenge banners, headers, dependency leaks, or version strings
  • field-notes.md - Long-form exploit notes: quick references for SQLi, XSS, LFI, JWT, SSTI, SSRF, command injection, XXE, deserialization, race conditions, auth bypass, and multi-stage chains
  • tools-mcp.md - MCP server integration for web: Playwright MCP (admin-bot DOM XSS reproduction, live DOM/console/network inspection, browser_network_requests lists endpoints but only browser_network_request part="response-body" returns payloads) and Burp Suite official MCP (27 tools, approval gates that block the agent on a modal dialog, target triple decoupling socket from Host: header, normalizeHttpContent rewriting CR/LF in the request prelude, map-based HTTP/2 that cannot express duplicate headers, send_to_intruder stages only, proxy history truncating items at 5000 chars, Collaborator payloadId/customData Pro-only); when to use each vs curl and ffuf
  • python-requests.md - Python requests toolkit: session scaffold, Burp-Intruder-like fuzzer (sync + ThreadPoolExecutor + httpx async), payload deploy from pat-reference.md wordlists, header/param spray, cookie/JWT, proxy
  • pat-reference.md — PayloadsAllTheThings index: bulk payloads for XSS/SQLi/SSRF/SSTI/LFI/Command Injection/Upload (requires PAT clone, see Prerequisites)

Read the full file on GitHub · 246 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday Changed · +1 lines c521b3a35bc9
  2. 8d ago Changed · +56 lines afdc2563b5d6
  3. 12d ago Changed · +3 lines e2c79ee49e1c
  4. 1mo ago Changed · +33 lines 20e8d2278208
  5. 1mo ago First seen · 153 lines · 124 tokens per session scan B 3fd4cafdb39b

Subscribe to this mod's changes

ctf-web is a skill published in the GitHub repository ljagiello/ctf-skills (3,549 stars, last pushed yesterday), licensed MIT. It adds 124 tokens to every session and 4,324 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

chief-data-officer-advisor

Chief Data Officer advisory for startups: AI training data rights and consent provenance, data product strategy (warehouse vs lakehouse vs mesh, build-vs-buy), B2B customer-data-as-asset valuation and M&A readiness, data team org evolution. Use when deciding whether to train models on customer data, choosing data…

alirezarezvani/claude-skills · 131 tokens

wrap-up

Close out a launched Claude Managed Agent — recap every primitive the founder now owns, regenerate the single-file overview page, and suggest the next 1-2 upgrades. Use when the user says "wrap up", "close this out", "what do I own now", "give me the summary", "recap the agent", or when the orchestrator routes…

alirezarezvani/claude-skills · 150 tokens

add-cloud-flow

Integrates Power Automate cloud flows into a Power Pages site. Lists available flows, suggests relevant ones based on intent, identifies scenarios and web roles, creates metadata files, and generates client-side code to call flows. Handles both new flow registration and adding already-registered flows to additional…

microsoft/power-platform-skills · 91 tokens

edit-site

Plans, coordinates and customizes PAC CLI-downloaded Standard or Enhanced data model non-SPA Power Pages sites created from Microsoft templates or manually in Design Studio. Use for requests involving pages, localized content, navigation, assets, snippets, site markers, Liquid/web templates, page layouts, or styling.…

microsoft/power-platform-skills · 104 tokens

deploy-site

Deploys an existing Power Pages site to a Power Pages environment using PAC CLI. Routes SPA sites to the existing build and upload-code-site workflow, and PAC-downloaded non-SPA (classic) sites to an isolated upload workflow with explicit data-model and website-identity checks. Handles tooling verification…

microsoft/power-platform-skills · 91 tokens

create-formal-verification-skill

Generate a project-local skill that formally verifies the code where a bug costs the most and tests reach the least, in any language, with whichever verification tool fits. Use for /create-formal-verification-skill, "model-check this race", "prove this invariant", or when concurrency, retries, state machines, parsers…

michael-denyer/pstack-claude · 82 tokens