Executes tasks iteratively until completion with structured status signals and stagnation detection.
npx skills add https://github.com/qf-studio/navigator --skill nav-loop-qf-studio
Please help me install this Agent Skill. Skill: nav-loop Source: https://github.com/qf-studio/navigator/tree/main/skills/nav-loop Command: npx skills add https://github.com/qf-studio/navigator --skill nav-loop-qf-studio
High-intent search queries and answers about installing and using this skill.
Invoke loop mode by saying "run until done" or "keep going until complete" followed by the task. The agent iterates through research, implementation, and verification phases, exiting only when completion indicators and an explicit exit signal are both satisfied.
Completion requires a dual-condition gate: at least two heuristics (code committed, tests passing, docs updated, marker created) plus an explicit EXIT_SIGNAL emitted as a pilot-signal JSON block. Either condition alone is insufficient to exit.
A stagnation detector hashes each iteration's state and triggers after a configurable threshold of identical states. By default it pauses for user input; with never_pause_on_stagnation enabled it auto-diversifies using combine, radical, or reread strategies.
Yes, set iteration_approval to "none" and never_pause_on_stagnation to true in .nav-config.json. A max_iterations cap is required as a safety guard so the autonomous loop cannot spin indefinitely on a broken task.
Avoid loop mode for single-step quick fixes, exploratory work without clear completion criteria, tasks blocked on external dependencies, and changes with security implications that require human review between steps.
brolagGenerate a structured todo.md with phases, tasks, and completion criteria.
SilentFlowerArchives completed Trellis tasks, records session journals, and commits bookkeeping changes.
dkyazzentwatwaDetect and diagnose quality degradation in human-AI collaboration conversations.
deerawanExecute small edits and configuration changes directly without plan files.
