Advance an agent loop one state transition and persist checkpoints.
npx skills add https://github.com/SollanSystems/loop-engineer --skill loop-run
Please help me install this Agent Skill. Skill: loop-run Source: https://github.com/SollanSystems/loop-engineer/tree/main/skills/loop-run Command: npx skills add https://github.com/SollanSystems/loop-engineer --skill loop-run
This Skill helps you operate a long-running agent loop without losing state, skipping verification, or pretending a task is finished before the gate proves it.
Use the loop-run skill to resume the existing loop workspace, advance exactly one transition, verify the result, and stop only when the next explicit terminal state is reached.
High-intent search queries and answers about installing and using this skill.
Resuming a long-running agent loop requires persisting checkpoints with state.json, task tracking, and run logs so the execution can recover safely and advance one state transition at a time without losing prior context.
State machine task orchestration for agent loops is the process of advancing execution through discrete states like planning, execution, and verification. It enforces independent gates and explicit terminal states to ensure tasks are only marked complete when verification proves it.
Handling approval pauses in an automated task workflow requires a state machine that persists the current checkpoint and halts execution. The workflow resumes from the approval state only after explicit human input is recorded and the next transition is triggered.
Yes, checkpointing can coordinate bounded repair after a verification failure by transitioning the state machine to a repair state. It persists the failure evidence and task status, allowing the agent loop to attempt fixes without re-planning the entire run.
Yes, operating this agent loop requires a skill-backed contract with state.json, task tracking, independent verification gates, and explicit seven-state termination. Without this structured contract, the loop cannot persist checkpoints or safely handle terminal states.
You should not use a single-step state transition approach when your workflow requires dynamic re-planning or lacks independent verification gates. This method is designed for bounded execution, repair, and approval pauses without re-planning the overall task strategy.
pazdedavPersist agent progress to 00-session-state.json and resume from checkpoints.
g5becksArchive completed project directories and restore them from a designated archive location.
XinAlohaOrchestrate autonomous agent loops with task sharding, isolated environments, and independent verification.
arthrodCoordinate multi-agent swarms with Claude Code's TeammateTool and Task system.
