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-windows-dll-hijacking-and-search-ordergit 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-windows-dll-hijacking-and-search-order)<a href="https://agentmods.dev/skills/unboundcompute/security-agent-skills/hunting-windows-dll-hijacking-and-search-order"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-windows-dll-hijacking-and-search-order/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-windows-dll-hijacking-and-search-order"><img src="https://agentmods.dev/badge/skills/unboundcompute/security-agent-skills/hunting-windows-dll-hijacking-and-search-order.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 | $0.00172 | $0.02041 |
| Opus 5.5 | $0.00069 | $0.00816 |
| Sonnet 5.5 | $0.00034 | $0.00408 |
| Haiku 4.5 | $0.00017 | $0.00204 |
Grade A, and why
hunting-windows-dll-hijacking-and-search-order 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 — 137 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hunting Windows DLL hijacking and search order: when a load-by-name lands in a writable directory
A Windows process that loads a library by name, rather than by a fixed full path, asks the loader to find it, and the loader walks an ordered list of directories. That is safe only when every directory the loader reaches before the real library is writable by no one below the process's own privilege. When a privileged process requests a library that is missing from its own directory, or when its application directory or a path entry is writable by a standard user, or when it uses a search mode that includes the current directory, the loader can resolve the name to a file the attacker planted. The privileged process then maps and runs that code as itself. The bug is not the load; it is a resolved search location that a lower-privileged user can write, upstream of where the genuine library lives. You find these by tracing each privileged load to the directories the loader searches and asking which of them a non-admin can write.
When to use
- A privileged executable or service loads libraries by name and you can observe its load behavior.
- The application directory, a path entry, or a working directory in the search order may be user-writable.
- A trusted signed binary can be copied to a writable directory alongside a planted dependency.
Scope check
Test library search order only on hosts you own or are authorized to assess, on non-production or a snapshot, using a benign marker library that proves it was loaded rather than one that takes real action. A confirmed hijack runs code inside a privileged process, so keep every probe within the authorized scope. If you can't name the authorization, stop.
The loop
- Establish where the loader actually resolves each library first. For each privileged load, determine the ordered directories the loader searches and whether the genuine library sits in a protected directory ahead of any writable one, or whether a missing library, a writable application directory, a writable path entry, or a current-directory search lets resolution reach a location a standard user controls. This is the false-positive killer: a load that always resolves from a directory writable only by administrators, ahead of every user-writable one, cannot be hijacked. Name the resolved location before crafting a plant.
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 · 137 lines · 172 tokens per session scan A b6ee10213d94
hunting-windows-dll-hijacking-and-search-order is a skill published in the GitHub repository UnboundCompute/security-agent-skills (5 stars, last pushed 1mo ago), licensed MIT. It adds 172 tokens to every session and 2,041 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 2026-09-09.
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.