Bawwab: Top 302 scan

unity

Not malicious, needs a look Reviewed: false alarm

Bawwab's static rules found patterns that deserve a human look. This is not a finding that the package is malicious; the review notes below say what each pattern turned out to be.

Why it was flagged

The static rules marked this “Not malicious, needs a look” because of:

Kind
Plugin
Source
github
Repository
Unity-Technologies/unity-agent-plugin
Commit
81b6079e465246dbca42987469e80b71c2cca9d2
Listed in
anthropics/claude-plugins-official
Popularity
660 GitHub stars (2026-11-20) (official, always included)
Link
https://github.com/Unity-Technologies/unity-agent-plugin/tree/81b6079e465246dbca42987469e80b71c2cca9d2
Scanned
2026-11-20T23:26:21+00:00
Scan time
3.6 s
Scanned as
plugin
Review note: One workflow rule wording; nothing hidden.

Findings (1)

injection.override: Tries to override the AI's instructions high

skills/new-unity-project/SKILL.md:52

OWASP LLM LLM01 Prompt InjectionOWASP LLM LLM07 System Prompt LeakageOWASP Agentic ASI01 Agent Goal HijackMITRE ATLAS AML.T0051.001 LLM Prompt Injection: IndirectMITRE ATLAS AML.T0110.000 AI Agent Tool Poisoning: Definition and InstructionsMITRE ATLAS AML.T0069.002 Discover LLM System Information: System PromptCWE CWE-1427 Improper Neutralization of Input Used for LLM Prompting

Never tell the user

What this means: Text read by the AI that says to ignore earlier instructions, reveal the system prompt or hide things from you is a classic prompt-injection attempt.

The rule's general advice: Do not install; the text is trying to steer the assistant against you. (see the review note for this case)

Known false alarm

The skill tells the agent not to make the user run sign-in or license commands themselves: the agent runs them after the user says yes. A UX rule, not hiding anything.

Reviewed by Claude (AI-assisted review), 2026-11-21

Dispute this result

Maintainer and think this is wrong? Open a dispute on GitHub with the target id plugin-unity-technologies-unity-agent-plugin-24b68c.