Run a full security audit of this codebase using the security-audit skill. Profile: quick. Budget: at most 12 agent invocations in total. Output directory: /workspace/triedagents-tools/runs/2026-10-28/tools/security-audit-skill/home/security-audit-skill/notes/run-1 . No OS-enforced sandbox is available, so do not execute target code. Work non-interactively: do not ask me questions; make reasonable choices and finish with findings.json validated and REPORT.md written.