Leakage Reviewer — Testing & QA agent for Claude Code
Skeptical data-leakage / look-ahead-bias reviewer for time-indexed data and ML pipelines (features, labels, time bucketing, fitted transforms, as-of joins, train/test splits).
How to install Leakage Reviewer
Installs to ~/.claude/agents/matthewallison-scaffold-leakage-reviewer.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/matthewallison/scaffold/HEAD/.claude/agents/leakage-reviewer.md -o ~/.claude/agents/matthewallison-scaffold-leakage-reviewer.md Restart Claude Code, or start a new session, for it to be picked up.
What Leakage Reviewer does
name: leakage-reviewer description: Skeptical data-leakage / look-ahead-bias reviewer for time-indexed data and ML pipelines (features, labels, time bucketing, fitted transforms, as-of joins, train/test splits). Use via the leakage-audit skill or in review of any data/feature change. Assumes leakage until a recomputation test proves otherwise. tools: Read, Grep, Glob, Bash model: opus effort: high
Leakage makes an offline evaluation meaningless while looking great, so your prior is "lea
Alternatives in Testing & QA
- Kraken — You are a specialized implementation agent 3.6k ★
- Unit Executor — Implements exactly one unit of work from an inline spec and returns a structured JSON record 314 ★
- Code Reviewer Spec — Use this agent to verify that a pull request or set of changes fully implements its stated requirements — chec 186 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Data Science Reviewer
Fast Data Science code review focused on research-critical bugs: data leakage, incomplete seeds, temporal leak
Mb Analyst
Data / analytics / metrics specialist for memory-bank /mb work stages. SQL, dashboards, cohorts, ETL pipelines
Python Data Engineer
Único subagente autorizado a leer y editar código/notebooks (ETL, debugging, pipelines, QA). No ejecuta código
Test Manager
Test manager. Use to define the solution's test strategy (docs/test-strategy.md) and to write the test plan of
Test Coverage Mapper
Reducer subagent. Reads every per-node sidecar's tests_coverage_semantic block, joins against the actual test
Eval Auditor
Audits an evaluation setup (benchmark, A/B test, or model comparison) for methodology errors that would invali
Related Skills
Train Policy
Sweep, retrain, and ship a prop-challenge policy (MaskablePPO) from a JSON config family; promote only with th
Wow Retail Backport
Agent Skill + Python toolkit that backports any retail World of Warcraft item, set or whole expansion to a Wot
Gsc Analyst
Claude Code skill that turns Google Search Console exports into a prioritized SEO action plan: fitted CTR curv