Retrospective Runner

The reason most retros produce platitudes is that they stop at the first answer. "Communication was bad" gets written on a sticky note, everyone nods, and the action item becomes "communicate better." That changes nothing because it doesn't address what was actually broken. This skill enforces the 5-Whys drill on every significant finding until you reach something structural — usually a process, template, tool, or norm — rather than a vague behavioural ask.

It starts with facts. Before opinions are allowed in the room, the skill builds a scorecard: planned vs actual on timeline, budget, scope, and the original success criteria. This prevents the retro from becoming a vibes-based reconstruction of what happened. With facts established, the 4L framework — Loved, Learned, Lacked, Longed For — gives structure to the discussion. Then the most important step: drill the top findings until you have a root cause that's actionable, not a symptom that's just observable.

Then the rule that separates real retros from ritual ones: experiments, not resolutions. Resolutions are forever and never reviewed. Experiments have a stop condition, a duration, a measure, and an owner. "We'll communicate better" is rejected. "For the next two sprints, we'll post a one-line decision log in #project-decisions after every meeting, owned by [name], and review at the next retro" is accepted. Maximum 3-5 experiments per retro — more than that and none get implemented.

Special protocols handle the failure modes. If the same issue appears in 3+ consecutive retros, previous experiments have failed and the skill escalates the approach — usually to structural solutions (mandatory written decision logs, tool changes, process changes) rather than behavioural asks. Psychological safety is handled explicitly: anonymous pre-collection for sensitive topics, leader speaks last (so they don't anchor the room), and an explicit "thank the critic" norm because punishing honesty in one retro kills honesty in all future ones. Not every project needs a formal retro — a 2-day task gets a 5-minute journal entry, not a 60-minute meeting. The skill scales the format to the work that was done.