Runs supervised multi-round DevOS Loop cycles to iteratively refine a target with human review between rounds.
npx skills add https://github.com/ruban-s/DevOS --skill loop-ruban-s
Please help me install this Agent Skill. Skill: Loop Source: https://github.com/ruban-s/DevOS/tree/main/skills/Loop Command: npx skills add https://github.com/ruban-s/DevOS --skill loop-ruban-s
--autoresearch flag enables autonomous iteration with a dead-ends ledger and MAD-based confidence gating on iteration scores. - Use Case: Point it at a prompt file with a goal like "more concise, less filler" and run 5 iterations, reviewing the output after each round before continuing. ## Quick Start Ask the agent to loop over a target file with a stated goal and an iteration count, for example: run /loop on prompts/summarize.md with the goal of making it more concise for 3 iterations.High-intent search queries and answers about installing and using this skill.
Invoke /loop with a --target path, an optional --goal describing what better means, and --iterations for the maximum number of cycles. Each round runs a full OBSERVE-to-LEARN cycle and carries learnings into the next round.
/loop runs full DevOS Loop cycles with human judgment threaded between rounds, while /optimize is an autonomous mutation loop with no human in it. Choose /loop when between-round review matters.
Yes, pass the --autoresearch flag to disable between-cycle human review. The loop then runs until the iteration count, target, or an interrupt, using a dead-ends ledger and MAD confidence gating on scores.
Use /loop --resume to continue a prior loop and /loop --status to show iteration history. Resumes read the dead-ends ledger first so rejected approaches stay rejected.
Each round is a complete cycle, so time and token cost is real and scales with iterations. Without a clear exit condition a loop can run indefinitely, and skipping the default human review removes the main feedback mechanism.
julianobarbosaRe-execute the Algorithm in loop mode to iteratively refine a target.
davduncExecute multiple Algorithm cycles with iteration-to-iteration learning and review.
pd-phucEnforces running verification commands before claiming work is complete or passing.
MozurokCritique draft engineering artifacts against a locked rubric and produce a revised draft.