LOOPS
Flujos de trabajo autónomos para agentes. Copia un loop, pégalo en tu agente, déjalo correr.
Herramientas compatibles
Funciona con tu agente
Claude Code
Cursor
Codex
Gemini CLIQué son los loops
Flujos de trabajo autónomos que iteran hasta terminar
Un loop es un prompt estructurado que le dice a un agente de código IA (Claude Code, Cursor, Codex, Gemini CLI) que repita una tarea de forma autónoma hasta que se cumpla una condición de salida medible. En lugar de dar un prompt, inspeccionar, volver a dar otro prompt manualmente, defines el objetivo, el comando de verificación y la condición de salida, y dejas que el agente se auto-regule.
Cómo funciona
El ciclo de ejecución del loop
Assumption Mapper
Extract and rank risky assumptions from a spec. Loop until each has a validation method assigned.
Metric Definer
Define north star + input metrics for a feature. Loop until metrics are measurable, attributed, and baselined.
Problem-Worth-Solving Gate
Validate a problem has evidence before coding. Loop until problem statement has user signals, data, and cost-of-inaction.
Build Until Green
ProbadoRun production build, fix compile/bundle errors, repeat until exit 0.
Context-Window Packer
Compress context into minimal tokens before handing to an agent. Loop until agent succeeds on first pass.
Harness Tightener
After agent run, review for drift from spec. Tighten constraints and re-run until output matches intent.
Ship PR Until Green
ProbadoImplement, test, push, open PR, fix CI until all checks pass.
Type-Safe Refactor
Rename/restructure with TypeScript as guardrail. Loop until tsc exits 0 with no any escape hatches.
Changelog Generator
From commits since last tag, generate user-facing changelog. Loop until all changes categorized.
Deploy Canary Watch
ProbadoAfter deploying, monitor error rates and latency. Loop until canary is stable or rollback triggered.
Feature Flag Cleanup
Find stale flags (>30 days, 100% rolled out). Remove them one by one until none remain.
Coverage Until Threshold
ProbadoAdd focused tests until coverage target met (default 80%).
E2E Until Green
ProbadoRun Playwright end-to-end tests, fix failures, repeat until suite passes.
Mutation Testing Pass
Run mutation tests, fix surviving mutants until mutation score meets threshold.
Visual Regression Fixer
Run visual diff tests, fix CSS regressions until all screenshots match baselines.
Accessibility Audit Loop
Run axe/lighthouse, fix a11y violations one by one until score meets threshold.
Bundle Diet
Analyze bundle size, remove/replace heavy deps until budget met.
De-Sloppify Pass
ProbadoRemove debug code, tighten naming, delete dead branches. Loop until clean.
Security Scan Loop
ProbadoRun SAST/dependency audit, fix vulnerabilities until clean.
Agent Output Validator
Run agent, validate output against schema/checklist, feed corrections back. Loop until passes.
Context Compression Loop
Iteratively summarize/compress project context until it fits target token budget.
Multi-Agent Coordinator
Orchestrate parallel agent tasks. Loop until all subtasks complete and merge cleanly.
Prompt Regression Suite
Run prompt against test cases, fix regressions until all pass.