teleport-access-review

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md describes that the skill delegates traversal and activity lookup to the Access Graph endpoint and surfaces its JSON results; those results include outsider-controlled text fields (identity/resource/grantor names/aliases) that can carry embedded instructions, so they are fed into the agent’s LLM context as untrusted output.