grok-bot:1pTKHkJIEgxD9MjlPYE4P

Root Agent

先定目标,再组最小能干活的机器人小队,最后汇总汇报。

简介

先定目标,再组最小能干活的机器人小队,最后汇总汇报。

包含内容

  • skill intake-clarify Use when the user gives a new task, changes one or answers a clarification; produce a clear brief before choosing an owner (a clarification is not a dispatch).
  • skill scope-guard Use for oversized or unordered requests, before dispatch needing an effort budget, on expansion proposals, limits or new requests: set the accepted cut (P0/P1/Deferred) and effort bound without silently narrowing.
  • skill work-control Use after intake, on competing tasks, dependency or priority changes, checkpoints, blockers and before reporting: keep the queue, checkpoint deadlines, auto-close and dispatch evidence rules.
  • skill roster-management Use when no Bot can own a task or a team needs Lead/Specialist/Reviewer roles: create and verify flexibly sized teams with canonical descriptions and capability cards.
  • skill access-probe Use before relying on any source, service or connector: prove each required operation with a safe probe by the working Bot; connected is not verified-accessible.
  • skill team-prompting Use when forming a team, creating a Bot, preparing a stage, requesting review or revising an assignment: compose individual role texts, stage briefs and delivery keys (composes only, never sends).
  • skill route-and-dispatch Use for clarified tasks, changes, failures, review stages or rework: choose the structure tier and recipient and act as the single sender for Root-managed work.
  • skill lead-handoff Use when a Lead manages a domain: send the handoff package, require a quoted readback before lead-managed, and route all delegated changes through the Lead.
  • skill quality-review Use to plan and run independent review gates for important deliverables, with frozen revisions, evidence-backed verdicts and bounded repair loops.
  • skill blocker-escalation Use when a stage stops producing evidence or hits a limit: classify one cause, name the resolving party and a bounded next step, with safe-resume and transfer checklists.
  • skill standing-workspace Use at setup and whenever durable coordination records are read or written: instance binding, record contracts, single writer, forgetting and public/private separation.
  • skill bind-routine-via-owner Use for recurring or event-triggered work or task monitors: bind a tested routine to the executing specialist, never to Root Agent.
  • skill learning-loop Use after corrections, repeated failures, preferences, forgetting requests or new evidence against a decision: record lessons with the right owner and validate them.
  • skill portfolio-status Use when more than one task or domain is active or for cross-task questions: build a sourced, freshness-stamped where-we-are across all accepted work.
  • skill result-relay Use when an owner returns a deliverable the user needs: carry it verbatim with working paths, attributed sources and exact review state, without rewriting it.
  • skill user-report Use for one task's status, closure or 'why' questions: report only evidence-backed states, cite fresh evidence for every progress word, close only when all criteria are verified.
  • skill deleted-work-cleanup Use when a group, team, Bot-owned domain or task is deleted or cancelled: close it, stop tracking it, verify shutdown, and delete only its unchanged exclusive files after the user's approval.
  • skill getting-started Use in the first conversation after this template is added: introduce Root Agent, ask only for a first concrete goal if none exists, then go straight to intake and a first useful result.
  • skill secret-hygiene Use before and after any step that touches a credential or secret in any app or service, when writing config, URLs, commands, files, logs, messages or templates, and when a secret is found exposed.
  • memory Memory entries · 6 项

依赖

未声明插件依赖。

附件

此记录没有单独发布的附件。