Skip to content
SkillModel-led5,291 tokens in use

hunt-cache-poison

A hunting guide for web cache poisoning and cache deception built from ten public bug bounty reports and named CDN research, for CDN-fronted applications.

14,523 stars on the repository MIT Licenseelementalsouls/Claude-BugHunter · skills/hunt-cache-poison
/plugin marketplace add elementalsouls/Claude-BugHunter /plugin install claude-bughunter@elementalsouls
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing hunt-cache-poison costs

171tokens idle
5,291tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

shelldetectednetworkdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Excellent
  • Frontmatter12/15
  • Structure25/25
  • Examples15/15
  • Output15/15
  • Size11/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent
  • Reaches the network−10

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 133
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 133

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity100
  • Maintenance84
  • Community68
  • Maturity28
  • Momentum40

hunt-cache-poison is a skill for Claude covering cache poisoning and web cache deception against CDN-fronted applications: unkeyed header injection such as X-Forwarded-Host, path-suffix tricks that get an authenticated page cached as a static asset, CDN allowlist gaps, and hop-by-hop header smuggling that lands a poisoned response at the edge. It separates this skill’s scope from hunt-host-header, which owns Host-header injection that reaches app logic without being cached.

The skill maps cache infrastructure signals, a ten-step method for finding unkeyed inputs and proving they are stored and served to other visitors, bypass techniques against WAF-blocked poison headers, and a blast-radius assessment. It names the public research behind each technique, including Omer Gil’s original web cache deception write-up and James Kettle’s 2024 path-normalization research, and cites ten disclosed cases from 2017 to 2024. It is for assets covered by a program or engagement you are authorized to test.

What You Get

  • A ten-step method covering cache-infrastructure mapping, unkeyed-header discovery, path-suffix web cache deception, DoS-via-cached-error testing, and blast-radius measurement
  • Request patterns for confirming caching behavior, forcing a cache miss for safe testing, and verifying a poisoned entry from a second client
  • A defense-to-bypass table covering WAF header-blocking, origin header stripping, and short-TTL cache purging
  • A three-question validation gate that requires the poisoned response to be reproduced from a separate client before it counts as a finding
  • Ten disclosed report citations from 2017 to 2024, including the original Omer Gil web cache deception write-up and James Kettle’s path-normalization research

Common Use Cases

  • Testing whether an unkeyed header like X-Forwarded-Host is reflected into a cached response on a CDN-fronted application
  • Checking whether appending a static file extension to an authenticated URL gets the full authenticated page cached and served to other visitors
  • Confirming whether a cached error response from a malformed header can deny service to legitimate users of a CDN-served asset

What Makes It Unique

The skill separates poisoning from deception explicitly and draws the boundary against the adjacent hunt-host-header skill, so the same Host-header injection technique is not reported twice under two different names. Its validation gate requires a second, unmodified request from a separate client or IP to show the poisoned response persisting in the cache, ruling out a self-cache that only the attacker’s own browser sees.

Under The Hood

The skill is a single SKILL.md of guidance with no bundled scripts. Following it, the agent sends HTTP requests with manipulated headers such as X-Forwarded-Host and crafted URL paths to cacheable endpoints, inspects response headers for cache-hit signals, and issues a follow-up request from a different context to confirm persistence. Use it only against assets you are authorized to test.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers