pragman-analyzelisted

# Pragman Analyze Produce a decision-grade analysis grounded in observable evidence. A good result preserves successful patterns, identifies causes without hindsight theater, and turns learning into explicit options—not silent configuration changes. ## Use this skill when - Work finished, stalled, or drifted and the next iteration should learn from what actually happened. - You have artifacts, diffs, tests, decisions, event aggregates, or carefully bounded user context to compare with the intended outcome. - You need explicit Keep/Change/Stop/Automate/Learn/Test next actions rather than a generic retrospective. ## Acceleration payoff It turns expensive rework into reusable operating learning while keeping confidence, evidence gaps, and trade-offs visible. ## Workflow 1. Define the subject, intended outcome, time boundary, systems and people in scope, and decision this analysis should enable. Distinguish successful, drifted, blocked, and incomplete work. For expanded work, classify each addition as necessary discovery, an approved scope change, or avoidable drift against the original decision boundary and contemporaneous decisions. If that evidence is missing, leave the classification unresolved. 2. Build an evidence index before interpreting. Prefer artifacts, diffs, tests, event aggregates, decisions, timestamps, and user-supplied context. Reference sensitive evidence by alias; do not copy private source bodies into the report. 3. Separate observed fact, sourced conte