backfilllisted

# /multiplai-context:backfill Reconstruct diary entries and learnings from existing Claude Code session transcripts. ## When to use - You've just installed multiplai-plugin and want to backfill from your session history - Sessions ran before the plugin was installed or before the diary pipeline was wired - You want to rebuild `now/` and `diary.json` catalog with historical depth ## Steps 1. **Scope the run** — decide the window: - Default: last 7 days (`--days 7`) - Extend: `--days 30`, `--since 2026-04-01`, or `--all` (entire history) - Scope to project: `--projects my-app,my-lib` 2. **Dry-run first** to see what will be processed: ``` uv run --project "${CLAUDE_PLUGIN_ROOT}/scripts" "${CLAUDE_PLUGIN_ROOT}/scripts/backfill.py" --dry-run [--days N] [--projects slug,...] ``` Review: session count, estimated tokens, and the **privacy notice** — backfill reads all local Claude Code transcripts. Use `--projects` to limit scope. 3. **Confirm with the user** — surface the dry-run output and ask: > "This will process N sessions (~X estimated tokens). Proceed?" 4. **Run the backfill** — long-running (minutes to hours, well past the Bash tool's 600s max timeout). You **MUST** invoke it via the Bash tool with `run_in_background: true`: ``` uv run --project "${CLAUDE_PLUGIN_ROOT}/scripts" "${CLAUDE_PLUGIN_ROOT}/scripts/backfill.py" [--days N] [--since DATE] [--all] [--projects slug,..