mit einem Klick
long-run-agent-harness
long-run-agent-harness enthält 6 gesammelte Skills von Tom-0727, mit Repository-Berufsabdeckung und Skill-Detailseiten auf SkillsMP.
Skills in diesem Repository
Use when designing or scaffolding a Codex-based long-running business agent, including deciding what belongs in AGENTS.md, skills, and subagents.
Create or modify hidden per-workdir heartbeat extensions for this harness. Use when asked to add pre-heartbeat prompt injection, post-heartbeat accounting/export/audit, or customize heartbeat lifecycle behavior for this deployed agent.
Use when work must be tracked across heartbeats, including durable multi-step todos, time-bound commitments, or scheduled reminders.
Default workflow for non-trivial iterative heartbeats. Use when advancing the assigned goal, continuing a task, handling a non-trivial human request, or doing work that needs an episode record and evaluator. Skip only for quick human responses, acknowledgements, status syncs, or other simple mailbox interactions.
On-demand workflow for software project development. Use when the task is to build or substantially rework a software system/module/tool. Drives an Assume → PRD → Plan state machine, each transition gated by human approval over mailbox.
Operate the mailbox system with three commands -> mailbox-send for sending messages, mailbox-read for reading messages, and mailbox-contacts for listing contacts.