Validate Sweep — Development skill for Claude Code
Autonomous validation loop — detect, validate, fix, re-validate until PASS or max attempts.
How to install Validate Sweep
Installs to ~/.claude/skills/krzemienski-validationforge-validate-sweep/SKILL.md
mkdir -p ~/.claude/skills/krzemienski-validationforge-validate-sweep && curl -fsSL https://raw.githubusercontent.com/krzemienski/validationforge/HEAD/commands/validate-sweep.md -o ~/.claude/skills/krzemienski-validationforge-validate-sweep/SKILL.md Restart Claude Code, or start a new session, for it to be picked up.
What Validate Sweep does
Autonomous validation loop — detect, validate, fix, re-validate until PASS or max attempts.
Alternatives in Development
- Ralphy — My Ralph Wiggum setup, an autonomous bash script that runs Claude Code, Codex, OpenCode, Cursor agent, Qwen & 2.8k ★
- Zeroshot — Runs coding agents as a graph: one agent implements, independent agents review, failures go to repair, and not 1.9k ★
- /validate — Run full validation on the current finding before writing a report 927 ★
Full documentation available on GitHub
View Source RepositoryRelated Skills
Validate Fix
Fix validation failures and re-validate until all journeys pass (3-strike limit).
Unity Ralph
Relentless verify-fix loop — refuses to stop until the project is clean. Runs unity-verifier repeatedly with c
i18n Validate
Validate translation files and FIX every problem found (missing keys, placeholder drift, plural errors), loopi
Forge Execute
Run validation journeys against the real system with autonomous fix loop
Evil Mode
EVIL mode — adversarial sweep with scored findings, adversarial verification, and loop-until-dry. The upgrade
Clear Rewrite
Rewrite text to pass CLEAR-100. Plain words, human voice, real structure, no filler. Runs the checker in a loo
Related Agents
Sweep Controller
Controls autonomous fix-and-revalidate loops until all journeys pass or max attempts exhausted
Content Director
Orchestrates the content-shorts pipeline 리서치팀 → 제작팀 ⇄ 법률팀 → 출시팀 for a single short (info/news/idol). Delegates
Fix Agent
Targeted fixes after task-verifier returns FAIL:fixable. Implements only what the verifier flagged; max 2 atte