journaling-sessionslisted

# Journaling Sessions Capture what a session produced as **structured, separable, retrieval-ready entries** — prose with metadata, one idea each — that a future session can find and reason about in isolation. The quality bar is not "does this summarize the session" but "will a future session, retrieving this cold months from now, be measurably better for having it." These entries are raw material: a downstream process clusters many of them into generalizations, and repeated reinforcement promotes the durable ones into long-lived guidance. Write each entry so that clustering works — that is what every rule here optimizes for. This skill produces raw capture. It runs an **automatic multi-pass loop** so you get thorough output from a single invocation — you do not need to ask for "multiple passes." ## What a session produces Three kinds of knowledge; capture all three, not only the most obvious: - **Conclusions** — decisions made, facts discovered, tradeoffs analyzed, hypotheses formed. What a meeting summary would contain. - **Process dynamics** — the shape of the reasoning: where breakthroughs happened, where we got stuck, what surprised us, what assumptions proved wrong. - **Perceptions** — what the user cared about, what engaged them or felt like overhead, what they pushed back on and why. The three-part framing is a coverage checklist, not a prediction of how entries cluster. Underjournaling conclusions is fine; underjournaling perceptions is a systematic blind