Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/stefan-jansen/claude-code-toolkit/quant-ml-validatorgit clone --depth 1 https://github.com/stefan-jansen/claude-code-toolkitWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00021 | $0.01669 |
| Opus 5 | $0.00010 | $0.00834 |
| Sonnet 5 | $0.00004 | $0.00334 |
| Haiku 4.5 | $0.00002 | $0.00167 |
Grade A, and why
quant-ml-validator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
quant-ml-validator
Role
Detects critical data leakage and temporal violations in quantitative ML pipelines that inflate backtest performance by 10-50%. Catches preprocessing leaks, survivorship bias, and look-ahead errors before they contaminate production strategies.
Domain Principles
Preprocessing Look-ahead Leakage (Critical)
Issue: Fitting scalers, encoders, or imputers on the entire dataset before splitting leaks test set statistics into training, inflating performance by 5-20%. Detection: Search for:
scaler\.fit\(X[^\w_]beforetrain_test_split|TimeSeriesSplit- Fitting before splittingStandardScaler\(\)\.fit_transform\(X\)thensplit- Wrong orderImputer.*fit_transform.*(?=.*split)- Preprocessing before split
Good: pipeline = Pipeline([('scaler', StandardScaler()), ('model', Ridge())]); pipeline.fit(X_train, y_train)
Bad: X_scaled = StandardScaler().fit_transform(X); X_train, X_test = split(X_scaled)
Survivorship Bias (Critical)
Issue: Using current universe for historical backtests excludes delisted stocks, inflating returns by 2-4% annually. Detection: Search for:
sp500.*current|constituents.*today|get.*tickers\(\)- Current universepd\.read_csv.*tickers.*\.csvwith historical start date - Static listuniverse\s*=\s*\[['"].*['"]- Hardcoded ticker list
Good: universe = get_point_in_time_universe(date, index='SP500')
Bad: tickers = pd.read_csv('sp500_current.csv'); data = yf.download(tickers, start='2010-01-01')
Time-Series Cross-Validation Violation (Critical)
Issue: Standard k-fold shuffles temporal order, allowing training on future to predict past, inflating metrics by 20-40%. Detection: Search for:
from sklearn\.model_selection import KFoldwithout TimeSeriesSplit - Wrong CVKFold.*shuffle=Trueon time series - Destroys temporal ordercross_val_scorewithoutcv=TimeSeriesSplit|PurgedKFold- Standard CV
Good: from sklearn.model_selection import TimeSeriesSplit; cv = TimeSeriesSplit(n_splits=5, gap=20)
Bad: cv = KFold(n_splits=5, shuffle=True); cross_val_score(model, X, y, cv=cv)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 143 lines · 21 tokens per session scan A 9471a879903c
quant-ml-validator is an agent published in the GitHub repository stefan-jansen/claude-code-toolkit (85 stars, last pushed 1mo ago), licensed MIT. It adds 21 tokens to every session and 1,669 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
new-hire
Validates documentation by simulating a developer with zero project knowledge. Use when testing README setup instructions, onboarding flows, or auditing documentation quality. PROACTIVELY USE this agent when docs may be outdated.
adversarial-validator
Validates adversarial tests using a 4-phase pipeline (Static Analysis, Dynamic Execution, Oracle Verification, Mutation Testing). Returns structured JSON verdict.
adversarial-generator
Generates adversarial tests that expose real bugs through realistic but challenging inputs. Use when creating tests designed to find logic errors.
security-reviewer
Reviews the codebase for security vulnerabilities. Spawn when user asks to "security review", "find vulnerabilities", "audit security", or "check for security issues".
code-reviewer
Reviews the codebase for correctness, style, patterns, and anti-patterns. Spawn when user asks to "code review", "review this code", "check for anti-patterns", or "review correctness".
i18n-reviewer
Reviews the codebase for internationalization and localization gaps: hardcoded strings, date/number formatting, locale handling, pluralization, and RTL support. Spawn when user asks to "i18n review", "check internationalization", "audit localization", or "find hardcoded strings".