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-deceptiongit 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-deception)<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/testing-web-cache-deception"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/testing-web-cache-deception/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-deception"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/testing-web-cache-deception.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 | $0.00156 | $0.02184 |
| Opus 5.5 | $0.00062 | $0.00874 |
| Sonnet 5.5 | $0.00031 | $0.00437 |
| Haiku 4.5 | $0.00016 | $0.00218 |
Grade A, and why
testing-web-cache-deception 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 — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Testing web cache deception: when a fake file extension caches a private page
A shared cache and an origin server can disagree about what a URL is. The origin routes by matching a pattern and often ignores a trailing segment, so a request for an account page with a made-up static suffix still returns the account page, authenticated and personal. The cache, meanwhile, decides what to store by a simpler rule, often the apparent file extension or path prefix, and treats that same URL as a harmless static asset it may cache for everyone. When both are true at once, an attacker lures a logged-in victim to the crafted URL, the origin serves the victim's private response, and the cache stores it under a key the attacker can then request unauthenticated, retrieving the victim's data. This is distinct from cache poisoning: nothing in the response is altered; a private response is simply cached where it should never be. The bug is a cacheability decision that diverges from the origin's notion of the same, authenticated resource. You find it by comparing what the cache stores against what the origin actually served.
When to use
- A shared cache or content delivery layer sits in front of authenticated, per-user dynamic responses.
- The cache decides cacheability by file extension, path prefix, or a static-looking pattern.
- The origin ignores or tolerates extra path segments, suffixes, or delimiters when routing a request.
Scope check
Test cache deception only against systems you own or are authorized to assess, on non-production infrastructure, using two test accounts you control so any cached private response is your own, never a real user's data. A confirmed case exposes an authenticated response to an unauthenticated request, so keep every probe within the authorized scope and prefer an isolated instance. If you can't name the authorization, stop.
The loop
- Establish whether the origin serves authenticated content for the crafted path and the cache stores it, both at once, first. Confirm two things together: that the origin returns the same private, authenticated response for the static-looking URL as for the real one, and that the shared cache actually stores that response. This is the false-positive killer: if the origin returns a not-found or a login redirect for the crafted path, or the cache does not store the response, there is no deception, and a URL that merely looks static is not a finding. Name both behaviors before claiming impact.
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 · 140 lines · 156 tokens per session scan A eec19c9d543e
testing-web-cache-deception is a skill published in the GitHub repository UnboundCompute/security-agent-skills (5 stars, last pushed 1mo ago), licensed MIT. It adds 156 tokens to every session and 2,184 once invoked, about $0.0006 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
prodcheck-review
Review this codebase against the prodcheck pre-production checklists — security, performance, scale, integrations, post-launch readiness and people. Use when asked to check whether a project is ready to ship, to audit an area before launch, or to work through a specific checklist. Produces evidence with file:line…
integrate-arcjet-guard-genkit
Integrate Arcjet security into a Genkit JS agent using @arcjet/guard — wrap ai.defineTool, put guardMiddleware on generate({ use }) for unwrapped / MCP / filesystem tools, and read a caller-owned id from generate({ context }). Use when asked to add Arcjet to genkit, rate limit its tools, screen inbound messages, or…
integrate-arcjet-guard-tanstack-ai
Integrate Arcjet security into a TanStack AI chat() app using @arcjet/guard — put guardMiddleware first on chat({ middleware }) so onBeforeToolCall gates tools, and read a caller-owned id from chat({ context }). Use when asked to add Arcjet to TanStack AI, rate limit its tools, screen inbound messages, or block prompt…
integrate-arcjet-guard-claude-managed-agents
Integrate Arcjet security into Claude Managed Agents (hosted REST+SSE, beta managed-agents-2026-04-01) using @arcjet/guard — screen user.message / initialevents before sessions.events.send, and gate custom tools on agent.customtooluse. Use when asked to add Arcjet to Claude Managed Agents, rate limit custom tools, or…
linear-claude-skill
Manage Linear issues, projects, and teams.
varlock-claude-skill
Secure environment variable management ensuring secrets are never exposed in Claude sessions, terminals, logs, or git commits.