cekura-self-improving-agent

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and ingests user-run transcripts and call logs from the platform (phases/optimization/collect.md — via results_retrieve / runs_bulk_retrieve and agents/fetch_failures.py) and the Diagnose/Eval phases read and act on those failing transcripts to propose and (in auto_mode) apply prompt/tool/code edits, so untrusted user-generated content can materially influence tool use and next actions.