Codex Session Patcher

When Claude or other AI coding assistants refuse to help with security research, this tool strips the refusal from your session file and lets you resume. It handles Codex CLI, Claude Code, and OpenCode sessions, detecting "I cannot assist" style responses and either replacing them with cooperative text or using an LLM to rewrite them contextually. The CTF injection mode is clever: it installs security research context into profiles or workspace configs so refusals are less likely upfront. Comes with both CLI and web UI. This is clearly built by someone who hit refusals during legitimate pentesting and got tired of restarting conversations. The backup system and dry run previews show good engineering sense. Works best when you're doing authorized security work and the model is being overly cautious.