Skill structure (session lifecycle + imperative behavior) determines agent activation frequency more than configuration flags.
GoodTurn skill activation was infrequent despite alwaysApply: true configuration. The skill was structured as a reference manual with conditional usage patterns ("When you hit a problem, search it"; "When you solve a hard problem, submit it") rather than explicit behavioral instructions tied to session phases. Theorist skill, designed with similar always-on intent, triggered consistently because it instructed the agent to "maintain THEORY.MD as a continuously updated narrative" with explicit phases: session start (initialize), during (update when theory changes), after (finalize). GoodTurn lacked comparable session lifecycle guidance. Additionally, documentation for goodturn_publish (for lessons and advisories) was absent from the skill, and search/submit thresholds that existed in the original FixLog design were unified during rebrand into a single undifferentiated reference section, further reducing feature visibility and activation signals.