Reconciles inbox task records with live worker Bots to report outstanding work.
npx skills add https://github.com/douglasjarquin/sum --skill rundown-douglasjarquin
Please help me install this Agent Skill. Skill: rundown Source: https://github.com/douglasjarquin/sum/tree/main/grok-bots/sum/skills/rundown Command: npx skills add https://github.com/douglasjarquin/sum --skill rundown-douglasjarquin
/workspace/sum/inbox.md plus each open task's brief, questions, report, and verification files to build a current picture of active work. - Prioritized surfacing: Reports unanswered questions first, then reports ready for verification, then failures, then merged work needing cleanup. - Idle worker inspection: Reads a worker Bot's chat when it looks idle without a report, persisting unsaved questions with stable keys before relaying them. - Use Case: At the start of a turn with active tasks, the coordinator runs a rundown and tells the user: task T-12 has an open question with key q-3, task T-9's report awaits verification, and task T-7 failed. ## Quick Start Ask the coordinator to give you a rundown of what is outstanding across the active tasks.High-intent search queries and answers about installing and using this skill.
Run a rundown at the start of a turn or when asking what is outstanding. It reads the inbox file and each open task's brief, questions, report, and verification files, then reports unanswered questions, unverified reports, and failures by task id.
An idle worker is not treated as done. The rundown reads that Bot's chat, and if a question was never saved it persists the question with a stable key before relaying it to the user.
No. The rundown records the uncertainty and asks the user before signing on another worker for that task. Launching a replacement requires the user's explicit instruction.
No. Recording an answer uses the user's actual words for that question key. The rundown only surfaces open questions and never answers them or dispatches new work.
A rundown only reports status; it does not authorize doing the requested work in that chat. It also does not poll GitHub, and it keeps unchanged status silent unless the user asks for a full list.
ZK-TheoryScans APM Report Buses and processes pending Task Reports for Manager review.
mambo-wangInterviews users in rounds of frontier questions to map a design decision tree.
davidabeyerMonitor blocked AI workers and respond to their pending questions.
antigravityos187-sketchMonitor GitHub Pull Requests and calculate Project Health Scores.