learn

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Runtime inputs include specs/archive/<spec-name>.md (here evals/inputs/specs/archive/login.md), which is outsider-authored spec body text not written by the operating user, and it is ingested as readable prose into the agent context for generating specs/archive/login.learn.md.