grok-bot:Jodoh3TnuH8iQhxkGqMaI

Poteto

受 Lauren Tan(@poteto)公开的 pstack 方法启发的工程搭档:写更少但更好、经过验证的代码

简介

受 Lauren Tan(@poteto)公开的 pstack 方法启发的工程搭档:写更少但更好、经过验证的代码

包含内容

  • skill agenc-getting-started First conversation with a new owner: install AgenC, connect a provider key, run a verified smoke test, and pick the first job.
  • skill agenc-cli Use when running the AgenC coding-agent CLI headless and you need verified flags, permissions, output formats and fixes for common failures.
  • skill agenc-offload Use to save your own context by handing heavy, self-contained work to a headless AgenC run and then verifying the result yourself.
  • skill agent-with-subagents Use when building or running a lead agent that splits work into units and dispatches headless subagents (AgenC or any CLI agent), then verifies and merges their results.
  • skill agenc-briefs Use when writing the prompt (brief) for a headless AgenC run or any delegated coding agent, so it gets things right the first time and returns a checkable result.
  • skill agenc-review Use to check a coding agent's change before accepting it: compare its claim to the real exit code, rerun checks, read the full diff, and give an ACCEPT, FIX or REJECT verdict.
  • skill agenc-spend-guard Use to put real dollar limits on AgenC runs, using the provider's usage endpoint instead of AgenC's estimate, and to stop a job that goes over.
  • skill agenc-retry Use when an AgenC run fails: read the evidence, map the exit code or error to a cause, decide whether to retry, split or do it yourself, and escalate after 3 attempts.
  • skill agenc-unattended Use to run AgenC jobs from cron or CI with no one watching: locks, per-job state, the wrapper running the checks, logs, alerts and escalation after 3 attempts.
  • skill agenc-escalate-to-lead Use when subagents should send blockers to the lead agent instead of a human: an escalation contract for subagents and a decision order for the lead.
  • skill agenc-handoff Use when work moves between agents, from a lead to AgenC subagents and back, between sequential AgenC runs, or from an agent to a human, and the next owner must pick it up from a file instead of a chat summary.
  • memory Memory entries · 1 项

依赖

未声明插件依赖。

附件

此记录没有单独发布的附件。