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 hunting-content-type-and-parser-confusiongit 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/hunting-content-type-and-parser-confusion)<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/hunting-content-type-and-parser-confusion"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-content-type-and-parser-confusion/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/hunting-content-type-and-parser-confusion"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-content-type-and-parser-confusion.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 | $0.00163 | $0.02163 |
| Opus 5.5 | $0.00065 | $0.00865 |
| Sonnet 5.5 | $0.00033 | $0.00433 |
| Haiku 4.5 | $0.00016 | $0.00216 |
Grade A, and why
hunting-content-type-and-parser-confusion 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 — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hunting content-type and parser confusion: when two parties disagree on what the bytes are
A stream of bytes has no inherent type; something decides whether it is HTML, JSON, an image, or a form. When that decision is ambiguous, the attacker gets to influence it. A browser that sniffs a response ignores a benign declared type and runs the HTML it finds inside. An upload filter that trusts the client type or the extension accepts a script disguised as an image. Two parsers on one path, a permissive multipart reader and a strict one, split a request differently and let content slip past the checker into the sink. A single file that is valid as two formats passes an image check and executes as markup. The bug is never the bytes; it is that the type is resolved by guessing rather than by an authoritative, enforced declaration. You find these by locating where content is typed and asking who decides and whether anyone can be made to disagree.
When to use
- User-controlled content is returned in a response with a missing, generic, or user-influenced type.
- Uploads are accepted and typed by client-supplied content type or by file extension rather than content.
- A request body or document can be parsed as more than one format along its path through the system.
Scope check
Test typing and parsing behavior only against applications you own or are authorized to assess, from test accounts, using benign markers rather than live payloads when you confirm that content is sniffed or a filter is bypassed. A confirmed stored case can execute in other users' sessions, so coordinate and stay in scope. If you can't name the authorization, stop.
The loop
- Establish the declared type and whether it is enforced first. For each response that echoes user content and each parser that consumes untrusted bytes, determine what type is declared and whether it is authoritative: is a correct content type and a nosniff response directive set, and does the consumer honor the declared type instead of guessing? This is the false-positive killer. A response with the right type and nosniff will not be sniffed as HTML, and a parser that binds to one declared format cannot be confused, so name the declared type and its enforcement before crafting anything.
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 · 139 lines · 163 tokens per session scan A 27a4659e6b0b
hunting-content-type-and-parser-confusion is a skill published in the GitHub repository UnboundCompute/security-agent-skills (5 stars, last pushed 1mo ago), licensed MIT. It adds 163 tokens to every session and 2,163 once invoked, about $0.0007 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
playwright-skill
IMPORTANT - Path Resolution: This skill can be installed in different locations (plugin system, manual installation, global, or project-specific). Before executing any commands, determine the skill directory based on where you loaded this SKILL.md file, and use that path in all commands below.
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…
analyzing-malicious-url-with-urlscan
URLScan.io is a free service for scanning and analyzing suspicious URLs. It captures screenshots, DOM content, HTTP transactions, JavaScript behavior, and network connections of web pages in an isolat.
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…