← ClaudeAtlas

session-statuslisted

This skill should be used when the user asks "what's up", "where are we", "catch me up", "what's the status", "what's left", "what should I do next", "remind me what we were doing", or runs /whats-up. Also applies when resuming a session after a break or a context compaction, before handing work to someone else, or when the user sounds lost about the current state. Briefs the user on where work stands — progress, in-flight work, remaining tasks, the single next step, blockers and suggested follow-ups.
pyang2045/claude-code-marketplace · ★ 0 · AI & Automation · score 60
Install: claude install-skill pyang2045/claude-code-marketplace
# Session status briefing Turn the current state of the work into something the user can absorb in five seconds. The failure mode this skill exists to prevent is **the confident summary from memory** — a recap that says "done" about something that was only discussed, or lists next steps that were superseded twenty messages ago. Gather evidence first. Write second. ## Step 1 — Gather evidence (before writing anything) Run these read-only checks in **one batched block**. All are best-effort: a failure here is a missing input, never a reason to stop. ```bash git branch --show-current 2>/dev/null git status --short 2>/dev/null git log --oneline -8 2>/dev/null git stash list 2>/dev/null ``` Then, still best-effort: ```bash gh pr status 2>/dev/null | head -30 ``` If the working directory is not a git repo, say so in one clause and continue — the conversation and todo list still carry the session. Then read the **todo/task list** if the session has one. It is the most direct statement of remaining work that exists; prefer it over inference. Only now revisit the **conversation**: decisions made, approaches abandoned and why, files touched, questions the user asked that were never answered. If the plugin's host session has an Obsidian session note or handoff in play *and it is already in context*, use it. Do not go hunting for one — this skill must work in any repo. ## Step 2 — Classify, and mark whatever cannot be proven Sort every item gathered into four buckets: | Buc