Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill attempts a comprehensive backtesting framework but suffers from significant verbosity—metric description tables and example output blocks explain concepts Claude already knows. The referenced scripts and files don't exist in the bundle, making the actionable commands non-executable. Nonsensical inline comments ('# 10000: 10 seconds in ms', '# HTTP 200 OK') suggest quality issues in the content generation.
Suggestions
Remove the three metric description tables (Performance, Risk, Trade Statistics) and the example output block—Claude already knows what Sharpe ratio and max drawdown are. This would cut ~50 lines.
Provide the actual script files (backtest.py, fetch_data.py, optimize.py, strategies.py, metrics.py) in the bundle so the CLI commands are executable rather than referencing non-existent code.
Add validation checkpoints: verify data was fetched successfully before backtesting, sanity-check results (e.g., warn if Sharpe > 5 suggesting overfitting), and validate parameter grid inputs.
Remove the nonsensical inline comments ('# 10000: 10 seconds in ms', '# HTTP 200 OK') which appear to be artifacts and add confusion.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is excessively verbose with large tables explaining metrics Claude already knows (Sharpe ratio, max drawdown, VaR, etc.), a full example output block, and descriptions of basic concepts. The metric description tables alone consume ~30 lines that add no value for Claude. The inline comments like '# 10000: 10 seconds in ms' and '# HTTP 200 OK' appear to be hallucinated/nonsensical artifacts that add confusion. | 1 / 3 |
Actionability | The skill provides concrete CLI commands with specific flags and parameters, which is good. However, no bundle files are provided, meaning none of the referenced scripts (backtest.py, fetch_data.py, optimize.py, etc.) actually exist, making the commands non-executable. The commands reference a framework that doesn't exist in the bundle. | 2 / 3 |
Workflow Clarity | Steps are numbered and sequenced (fetch data → run backtest → analyze results → optimize), but there are no validation checkpoints or error recovery steps. For a financial backtesting workflow where bad data or misconfigured parameters could produce misleading results, there should be data validation steps and result sanity checks. | 2 / 3 |
Progressive Disclosure | The skill references external files (errors.md, examples.md) and has a clear file structure table, which is good progressive disclosure design. However, no bundle files are provided, so all references point to non-existent files. The main SKILL.md also includes too much inline content (metric tables, example output) that should be in reference files. | 2 / 3 |
Total | 7 / 12 Passed |