retrospectiva

Run the retrospective skill after completing a delivery increment. It first records the cycle with predicted versus actual ICE scores and effort, then walks through what went well, what didn't, root cause analysis, and BVSSH dimension checks.

This skill combines fishbone (Ishikawa) diagrams for breadth across people, process, product, platform, principles, and pressures, with 5 Whys for depth on the top-ranked cause. It also maps problems to Ohno's 7 wastes before analysis.

The skill records predicted ICE scores, effort estimates, and risk assessments alongside actual outcomes in a cycle history file. Calibration summaries accumulate across cycles, prompting a framework health review every 5 cycles.

Yes, the skill supports standalone mode where canvas and persistence references are treated as optional. It delivers the retrospective artifact directly in the conversation and only writes to authorized canonical sources when requested.

The skill requires drafting counter-arguments for each major claim and checking whether 'should have seen it' claims were actually predictable from pre-outcome evidence. This prevents manufacturing false patterns from post-hoc rationalization.