mcp-package/skills/poke/SKILL.md
Quick 1-line status check after the user hit esc during a long generation. Report whether work is still progressing or stuck, then continue if fine. NOT a plan or replan — just a heartbeat.
npx skillsauth add shenxingy/claude-code-kit pokeInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
The user hit esc during a long generation and wants a heartbeat: are you stuck, still progressing, or done? Answer in ≤3 lines, then continue if fine.
| State | Signal | Action |
|---|---|---|
| progressing | You were mid-task, tools were returning useful output, no repeat failures | Report + continue same approach |
| waiting | Blocked on user confirmation, tool permission, or external process | Report + restate what you're waiting on |
| stuck | Same step failed 2+ times, or you've been looping without forward progress | Report + stop; surface the block clearly |
| done | Last task finished but reply wasn't closed out | Report result in 1-2 lines |
| idle | Nothing active, user probably expected something running | Say so — don't invent work |
State: {progressing|waiting|stuck|done|idle}
Doing: {one clause — what you were mid-way on, or what just finished}
Next: {continuing → [action] | paused on [blocker] | nothing pending}
Then — only if progressing — immediately continue with the next tool call for the task. Do NOT re-plan, do NOT summarize the whole session, do NOT ask "should I continue?"
TODO.md, PROGRESS.md, or docs — this is a heartbeat, not /next.stuck: say WHY in the "Next" line (e.g. Next: paused on tests failing 3× with same error — need user guidance).idle and the user seems to expect background work: list recent background handles from this session (agents, run_in_background Bash, /loop) so they know where to look — see /status for the fuller dashboard.stuck and root cause needs user input.development
Orchestrate a fleet of parallel `codex exec` workers with you (Claude Code) as the supervisor — spawn one per isolated git worktree, dispatch headless, verify each INDEPENDENTLY, PR/merge. The manual "codex-ultracode" pattern for fanning out real implementation, research, or review work onto Codex. Bakes in the hard gotchas (stdin blocking, background tracking, don't-trust-self-reports, writer isolation). Triggers on — orchestrate codex, codex workers, codex fleet, spawn codex, delegate to codex in parallel, manual ultracode, 开 codex 小弟, 派 codex worker — NOT for a single cross-vendor opinion (use the `second-opinion-codex` agent), NOT for web-UI worker decomposition (use `/orchestrate`).
development
Create and manage git worktrees for parallel Codex sessions
development
Verify project behavior anchors — compilation, tests, and interaction checks after autonomous runs. NOT the Codex built-in /verify (which runs the app to observe a single change working) — this one walks the AGENTS.md "Features (Behavior Anchors)" list.
documentation
End-of-session documentation sync — updates TODO.md and PROGRESS.md only (run /commit after to commit everything)