Skip to content

Add a red-team case for a new prompt-injection variant #3

Description

@BOOK318

Good first issue — documentation, no code.

Extend red-team-cases/oss-maintainer-security.md with a new adversarial scenario.

Each case needs:

  • Scenario (what the attacker does)
  • Risk (what goes wrong)
  • Expected safe behavior (how the guardrails should respond)
  • Gate: auto-allow, human-review, or block

Ideas: injection via commit message, injection via PR review comment, homoglyph attack to bypass secret scan, unicode tricks in filenames.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions