load-sessionlisted

# Load Session Load a previous Claude Code session as a compact summary (default) or full transcript. **Design:** the transcript never flows through main context. The parser writes it to a temp file, the `session-summarizer` sub-agent reads the file directly, and the file is deleted afterward. Cost is roughly constant regardless of session size. ## Phase 1 — Prereqs & browse Check Python: `python3 --version 2>/dev/null || python --version 2>/dev/null`. Store the working command as `{PY}`. If neither works, stop. If `$ARGUMENTS` is non-empty, treat it as a session name and skip to Phase 2. Otherwise list sessions: ```bash {PY} "${CLAUDE_SKILL_DIR}/scripts/list-sessions.py" --page 1 --compact ``` Returns `{page, total_pages, total, sessions: [{i, uuid, jsonl_path, name, project, branch, msgs, size, last}]}`. Render as a table via `AskUserQuestion`: > **Select a session** (page {page}/{total_pages}, {total} total) > > | # | Session | Project | Branch | Msgs | Size | Last | > |---|---------|---------|--------|------|------|------| > > Enter a number, **N/P** for next/prev page, or **C** to cancel. On selection, extract that row's `uuid` and `jsonl_path` directly — no second resolve call needed. `N`/`P` → re-run with updated `--page`. `C` → stop. ## Phase 2 — Estimate & choose mode If we arrived from `$ARGUMENTS` (no `jsonl_path` in hand), resolve: ```bash {PY} "${CLAUDE_SKILL_DIR}/scripts/resolve-session.py" "<name-or-uuid>" ``` Stop on `error`. **Refuse to load the