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.
npx skills add UnboundCompute/security-agent-skills --skill testing-web-cache-attacksgit clone --depth 1 https://github.com/UnboundCompute/security-agent-skillsWrote 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.
[](https://agentmods.dev/skills/unboundcompute/security-agent-skills/testing-web-cache-attacks)<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/testing-web-cache-attacks"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/testing-web-cache-attacks/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.
<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/testing-web-cache-attacks"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/testing-web-cache-attacks.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00113 | $0.01452 |
| Opus 5.5 | $0.00045 | $0.00581 |
| Sonnet 5.5 | $0.00023 | $0.00290 |
| Haiku 4.5 | $0.00011 | $0.00145 |
Grade A, and why
testing-web-cache-attacks scanned grade A with 0 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 1mo 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Testing web cache attacks: the bug is the key, not the app
A shared cache trades correctness for speed by serving one stored response to many users, and that sharing is the vulnerability. If an attacker can influence what gets stored (poisoning), everyone downstream receives their payload; if an attacker can make the cache store someone else's private response under a key the attacker controls (deception), they read data that was never theirs. Both live in the gap between what the cache keys on and what actually determines the response.
When to use
- You are reviewing a CDN, a reverse proxy, or any shared HTTP cache in front of an app.
- Responses that depend on headers or user identity are served through a cache.
- You are assessing whether authenticated content can be cached and re-served.
Scope check
Test caches and apps you own or are authorized to test. Use benign markers and your own accounts; do not poison responses served to real users or read real users' data. If you can't name the authorization, stop.
The loop
-
Map the cache and its key. Identify the caching layer in front of the app and determine the cache key: which parts of the request (path, query, some headers) decide whether two requests share a stored response. Everything the response depends on but the key ignores is an unkeyed input, and unkeyed inputs are where both attacks live.
-
Find what is cacheable. Determine which responses the cache stores and for how long: static-looking paths, extensions, explicit cache headers, and any response the cache decides to keep. A response that is cached but depends on request data the key omits is a poisoning candidate.
-
Test cache poisoning through unkeyed input. Send a request whose unkeyed header or parameter changes the response (a reflected header, a header that sets a link or script origin, a header that triggers an error page), then confirm the poisoned response is stored and served to a normal request without your input. If a later clean request receives your payload, the cache is poisoned for every user of that key.
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.
- 1mo ago First seen · 124 lines · 113 tokens per session scan A 17b8c9b8c4ff
testing-web-cache-attacks is a skill published in the GitHub repository UnboundCompute/security-agent-skills (5 stars, last pushed 1mo ago), licensed MIT. It adds 113 tokens to every session and 1,452 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.
Other skills, from other repositories
integrate-arcjet-guard-cloudflare-think
Integrate Arcjet security into a Cloudflare Think app using @arcjet/guard — assign guardHooks().beforeToolCall on a Think subclass so the hook gates tools, and read a caller-owned id from helper options. Use when asked to add Arcjet to Cloudflare Think, @cloudflare/think, rate limit its tools, screen inbound messages…
protect
Add Arcjet request protection to JavaScript and TypeScript HTTP handlers. Use when protecting Next.js, Express, Fastify, SvelteKit, Remix, React Router, Hono, Astro, Nuxt, Bun, Deno, NestJS, or Node.js routes with rate limiting, bot detection, Shield, email validation, or sensitive info detection.
create-auth-skill
Scaffold and implement authentication in TypeScript/JavaScript apps using Better Auth. Detect frameworks, configure database adapters, set up route handlers, add OAuth providers, and create auth UI pages. Use when users want to add login, sign-up, or authentication to a new or existing project with Better Auth.
fhir-developer-skill
FHIR API development guide for building healthcare endpoints. Use when: (1) Creating FHIR REST endpoints (Patient, Observation, Encounter, Condition, MedicationRequest), (2) Validating FHIR resources and returning proper HTTP status codes and error responses, (3) Implementing SMART on FHIR authorization and OAuth…
api-security
Reviews REST and GraphQL APIs against the OWASP API Security Top 10:2023. Auto-invoked when reviewing OpenAPI/Swagger specs, API endpoint code, or GraphQL schemas. Covers BOLA, BFLA, authentication, rate limiting, and SSRF. Produces findings mapped to API1-API10 with remediation guidance.
create-auth-skill
Scaffold and implement authentication in TypeScript/JavaScript apps using Better Auth. Detect frameworks, configure database adapters, set up route handlers, add OAuth providers, and create auth UI pages. Use when users want to add login, sign-up, or authentication to a new or existing project with Better Auth.