From harness-ops
Run a task as a supervised verification loop — 3 gates (Pass/Fail, Quantitative, Qualitative), iterate until they pass, escalate at autonomy boundaries
How this command is triggered — by the user, by Claude, or both
Slash command
/harness-ops:loop [task or goal]This command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
Read `${CLAUDE_PLUGIN_ROOT}/skills/loop/SKILL.md` and follow its instructions exactly.
User arguments: $ARGUMENTSRead ${CLAUDE_PLUGIN_ROOT}/skills/loop/SKILL.md and follow its instructions exactly.
User arguments: $ARGUMENTS
npx claudepluginhub hyunho058/harness-ops --plugin harness-ops/loopExecutes tasks in iterative loops with configurable max iterations, exit conditions, stall detection, and progress tracking. Also supports metric-driven optimization with automatic revert on regression.
/loopParses natural language task into components (task, max iterations, completion condition) and sets up repeating AI execution loop via bash script.
/loopRuns exec-verify-fix loops until acceptance criteria pass or blockers identified, producing loop status, evidence table, remaining todos, and next command.
/loopLaunches a bounded controller service that orchestrates four-phase child runs (plan, execute, validate, reflect). Supports attach, resume, watch, and stop operations via flags.
/loopConfigures and launches a Ralph headless LLM harness loop that runs code generation, tests, builds, spec checks, and audit gates with Shen backpressure. Requires /sb:init first.
/loopBenchmarks and evolves prompt assets like skills and commands in isolated branches using loop suites. Supports list-suites, show-suite, run with targets/budget/parallel/generations, and resume subcommands.