dream-rememberlisted

# Multiplai: Process Learnings Human-in-the-loop workflow for applying accumulated session learnings to memory files. Dream (nightly or on demand via `/multiplai-context:dream`) generates a proposal file in `.multiplai/dreams/`. This skill loads that proposal, walks through it with the user, and applies approved changes. ### The proposal file is the decision record The proposal `.md` is itself the record of what has been reviewed. When an item is decided — applied, edited, or rejected — its block is **moved into a `## Processed` section at the end of the file** (via `dream.py --mark-processed`, never by hand). Items under `## Processed` are **no longer pending**: this skill (and any future client) skips them, so a proposal can be reviewed in more than one sitting or tool with no double-applies. That one heading is the entire cross-tool contract — there is no sidecar file. A proposal is archived (Step 6) only once **nothing is left pending**. --- ## Step 1: Locate the Proposal Check `.multiplai/dreams/` for a file matching `processed-learnings-*.md` — top level only, skip the `applied/` and `rejected/` subdirectories (those hold already-reviewed proposals; never recurse into them). Take the **most recently modified** (newest mtime — not lexical name order; a same-day re-run writes a `-2`, `-3`, … suffixed file that is newer but sorts *before* the base name). - **Found:** load it, report its date and summary line to the user, and **record its exact path — Step 6 archi