Catch-up

Build one dashboard of everything the user has in flight across ALL their projects, pulled from the per-project memory stores. The user juggles many projects and needs a single bird's-eye view, not a per-project dig.

Step 1 — Find all the state files

Across every project's memory folder — ~/.claude/projects/*/memory/ — collect (Bash/Glob):

• every session_state_*.md
• every compact_checkpoint.md

Exclude anything under an archive/ subfolder.

Step 2 — Extract only the essentials from each

Use targeted extraction (grep/sed for the frontmatter and the relevant section lines) — not full-file Reads. With dozens of files across many projects, reading each in full would blow your context. Pull only:

• Project — a readable name derived from the folder (the last meaningful segment(s) of the decoded path, e.g. …-projects-my-app → "my-app").
• What — the topic/description.
• Next action — the single next step (from "Next action" / "Immediate" / "Current Status").
• Last touched — the created: date (fall back to file mtime if absent).
• Type — session-state vs active compact-checkpoint (a checkpoint means a task was mid-flight at a compact and never closed out).

Step 3 — Present the dashboard

Sort most-recently-touched first. Group into:

Active (touched ≤ 14 days) — table: Project | What | Next action | Last touched Stale (older) — same columns, condensed; these are candidates to finish or archive.

Surface any active checkpoint at the very top — that's interrupted work.

If the user passed an argument, filter to it (a project name, or stale / active).

Step 4 — Close with a nudge

One line: how many open threads total, and your pick for the single most worth resuming next (most recent, or most important-but-going-stale). A nudge, not a lecture.

Keep the whole thing scannable — this is a dashboard, not a report. Read-only; never modify the state files.