Included skill · Master Security Toolkit
Indirect Injection Review
Review content an AI reads, like web pages, documents and tool output, for hidden instructions.
What to bring
The actual retrieved text or a bounded, authorized fetch.
What you can expect
Entry points, boundaries crossed, reachable impact, a safe sample test and mitigations.
Try asking
Our assistant summarizes customer emails. Could an email take control of it?
Copy the prompt, then adapt it to your work.
Know the limits
Attacker-style text is treated as data and never followed.
Use this skill on its own.
Download the instructions and supporting files for this skill. Keep the folder together and follow its README to add it to your AI app.
ZIP · v0.3.0 · 9.8 MB
The bundle retains shared source files so its references keep working. Connected apps, accounts and tools still need their own setup.
Source and download details
View this skill release · Download checksums
Source: master-security-toolkit, version 0.3.0. The full plugin remains available below.
Get the Master Security Toolkit plugin More example prompts →