Skip to content

Security: prompt-injection guidance for adapters #143

Description

@himanshu231204

Background

Model output is untrusted but may be treated as safe.

Why it matters

Prompt injection can mislead tooling.

Acceptance criteria

  • Document that model output is untrusted; never execute it.
  • Add to SECURITY.md.

Where to look

  • SECURITY.md

How to claim

Comment /claim on this issue.

Metadata

Metadata

Labels

Type

No type

Fields

No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions