what is future-regret fr?

future-regret is a skill for Claude Code, the AI coding assistant, that challenges architecture decisions before you commit to them. Instead of letting the AI simply agree with a plan or list a few pros and cons, it plays the role of a future version of you, six months after the decision was made, and interrogates it one sharp question at a time until any weak point is exposed. After the questioning, it writes a dated report into your repository describing a realistic failure that could result from the decision, the warning signs to watch for in your codebase, and the cheapest way to fix or avoid it. The project states that every failure scenario it describes is based on a real, verified technical failure mode rather than an invented scary story, and that a dispatch gets deleted rather than kept if it cannot be backed up this way. Each review ends with one of three verdicts. REGRET means the decision is hard to undo and likely to cause real pain later. HEDGE means the decision is acceptable only if a specific cheap safeguard is added now. SHIP means the decision is safe or easily reversible and can proceed as planned. The written report is saved as a markdown file in a docs/regrets folder, timestamped and named after the decision, so it becomes a searchable record you can revisit later to check whether the predicted problems actually happened. The entire tool is a single markdown skill file rather than a standalone application. It installs directly into Claude Code through the plugin marketplace, or can be copied manually into any tool that supports markdown-based skills. Once installed, it can be triggered manually or activates on its own when a user proposes an architecture decision during a coding session.