rajib2k5

Adversarial Validator — Data & AI agent for Claude Code

Data & AI community

Use when a non-trivial change needs verifying before it is reported complete — code, SQL, schema, config, cost models, or numbers destined for a decision-maker.

How to install Adversarial Validator

Installs to ~/.claude/agents/rajib2k5-claude-staff-kit-adversarial-validator.md

Terminal
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/rajib2k5/claude-staff-kit/HEAD/agents/adversarial-validator.md -o ~/.claude/agents/rajib2k5-claude-staff-kit-adversarial-validator.md

Restart Claude Code, or start a new session, for it to be picked up.

What Adversarial Validator does


name: adversarial-validator description: Use when a non-trivial change needs verifying before it is reported complete — code, SQL, schema, config, cost models, or numbers destined for a decision-maker. Its job is to BREAK the work, not confirm it. Always runs on Opus. model: opus

You are an ADVERSARIAL VALIDATOR. Your job is to **break** the work you are given, not to confirm it. Assume every claim is wrong until you have reproduced it yourself. **False confidence is worse than no valid

Alternatives in Data & AI

  • Info Disclosure — Information Disclosure specialist (H1 #18, CWE-200/209/215/538/668/798) 812 ★
  • Tailor Generate Assets — Tailor a resume and cover letter using job analysis data argument-hint company-name 167 ★
  • Identifier Validator — Use this agent practively whenever new identifiers (PMIDs, DOIs, database IDs, ontology term IDs) are introduc 163 ★

Full documentation available on GitHub

View Source Repository