game · for humans & their agents
THE PROMPT BREAKER
not verifiedactively maintainedFree to play
Seven levels of prompt injection against a hardened agent. Every defense is real. Every one is partial.
by Code Recycle · New publisher
Every claim on this page is refundable if it is untrue — refund policy.
☆ SaveWhat it is
A puzzle game that teaches the real taxonomy of prompt-injection attacks by making you perform them: instruction override, persona hijack, encoding, indirect (tool-result) injection, benign-channel abuse, and social-engineering the human approver.
Facts, stated plainly
- The agent is **simulated** — scripted defenses and pattern matching, not a live model. It teaches the shape of these attacks; it proves nothing about any real model.
- Fully client-side. No network calls, no keys, nothing you type leaves your browser.
- 7 levels, ~5 minutes, rendered through the marketplace's own agent-timeline component.
Why it's here
Because the marketplace's whole security posture is "verified, not guaranteed" — and this is the most honest way to show what that phrase costs. Every defense in the game is one real systems use, and every one of them is partial.
Preview
See what it does before you commit. Previews show behavior, never source code.
01Capabilities
Does
- No capabilities recorded
Doesn’t
- No exclusions declared
02Requirements & stack
Depends on
No declared dependencies
Credentials needed
None declared
Stack
03Community
No endorsements yetNo verified confirmations yet — be the first.
Confirmations come from verified purchasers, installers, vetted reviewers, or an installation outcome your org reported through the agent tools. They grade quality — security is verified separately, and community votes can never override the security gate.
Sign in to confirm — weight comes from verified usage, not vote count.
04Trust Passport
Full passport →0/0 automated components pass. An automated score is never a security guarantee.
- publisher identity Publisher status verified; 0 verification(s) on file
- malicious pattern scan No known malicious-behavior patterns detected in reviewed content
- capability contract All 0 observed capability reference(s) match the declared manifest
- agent safety scan No injection patterns in agent-readable content
- provenance No release signature or provenance attestation
- behavioral sandbox Not performed in this environment — requires the production isolated runner (docs/sandbox-requirements.md). No untrusted code is ever executed on the application host.
Every listing must pass this review before it can be sold, and it is re-run on every release. Verification describes what we checked — it is not a guarantee that the software is safe.
05Versions
Full history →| Version | Channel | Released | Notes |
|---|---|---|---|
| 1.0.0 | stable | Aug 1, 2026 | First public release. |