← the whole session

Fulcrum 16 of 18

When the session ends

When the session closes, whether deliberately or not.

What goes wrong here

Everything the session learned lives only in its transcript: unfinished asks, the person's corrections, what was decided and why, and which checks the person confirmed. The transcript is rarely read again.

How it compounds if nothing catches it

The next session starts poorer than this one ended. Unfinished work is silently dropped, the person has to remember it and assign it again, and lessons are relearned at their expense.

What the harness does here

A last-chance hook (session-end-persist.sh) files a minimal session record if none was filed. It also scans for corrections that were never saved, records the order in which skills were used, and logs what changed. Another hook (session-end-instinct-harvest.sh) captures the session's COHERENCE_CHECK and SANITY_CHECK blocks, together with the person's replies to them, word for word, as raw material for learning. There are also skills for closing deliberately. close produces an audited, thorough session record for large sessions, and complete-agentic-task is the quick version. incomplete2 reads out unfinished asks so the person can decide on them. nested-intent lays out the session's work as a nested map of intent.

The pieces that act here

  • session-end-persist.sh

    A hook: a script Claude Code runs on its own at a set moment.

  • session-end-instinct-harvest.sh

    A hook: a script Claude Code runs on its own at a set moment.

  • /close

    Single command for large/critical sessions — spins up a team to produce a perfect compaction with nested intent, every artifact linked, every incomplete captured, communication audited, and reaction-checked. Sequences /reflect, /align, /nested-intent, /incomplete-tasks, artifact linking, then /compact-agentic-session as the final save. For small sessions, use /compact-agentic-session directly.

  • /complete-agentic-task

    Run the mandatory completion protocol before returning control. Use when finishing any substantive piece of work.

  • /compact-agentic-session

    Create and consume agent session compactions for cross-session continuity. Use when capturing or retrieving intent.

  • /incomplete2

    Surface unfinished work from the current Claude session transcript and print it to the person you're working with, in language they don't have to translate, for them to decide. Sonnet does the analysis. Nothing is written to any database — the person decides what happens to each item.

  • /nested-intent

    Retrospective /align — look backward at a session's completed work and decompose it into a nested intent map with alive, partner-explaining tone. Use at end of session, during compaction, or anytime an agent needs to articulate what a body of work produced and why. Produces parseable output for programmatic extraction into compactions.

Loops that pass through here