Helix Archive

The session already happened — here, in this conversation. Your job is to curate it into a story someone can read in five minutes: what was tried, what was decided and why, what came out of it. The log is append-only: an archive is a record, not a workspace. Nobody answers anything in it; the one thing a viewer can do is promote a memory candidate.

Boundary (ratified): Stream = watching work as it runs · Archive = the story written afterwards. If the user wants to watch live, that is a stream; archive it when it's over.

Setup

# what level is this archive at, and where did the last snapshot stop?
# reads the log directly; works with no server
${CLAUDE_PLUGIN_ROOT}/bin/helix archive info --archive <slug> [--dir <projectDir>]