Content
78%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable skill body: every instruction is an executable command with real flags, errors have cause/solution mapping, and detail is properly deferred to real, clearly-labeled reference files one level deep. The main weakness is token waste from the Examples section duplicating commands already shown in Instructions, plus a missing explicit setup-verification step.
Suggestions
Remove the three commands duplicated verbatim between Instructions and Examples — replace the Examples section with distinct scenarios or a one-line pointer to references/examples.md to reclaim ~20 lines of token budget.
Drop the Contents nav line at the top of the body; it duplicates the section headers that follow immediately.
Add a quick setup-validation step to Prerequisites (e.g., a single smoke-test command confirming the RPC connection works) before users launch into full simulations.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids concept explanations and is mostly commands, but the Examples section repeats three commands verbatim from the Instructions section ("python ${CLAUDE_SKILL_DIR}/scripts/flash_simulator.py arbitrage ETH USDC 100 --dex-buy uniswap --dex-sell sushiswap" appears twice in full), and the Contents nav line duplicates the section headers. This is more than minor trimming — roughly 20 lines of duplicated content — matching the "mostly efficient but could be tightened" anchor rather than 4. | 3 / 5 |
Actionability | Every instruction is a copy-paste-ready executable command with concrete arguments and flags ("arbitrage ETH USDC 100 --dex-buy uniswap --dex-sell sushiswap", "liquidation --protocol aave --health-factor 0.95", "--compare-providers", "--risk-analysis", "--full --output json"), covering the common cases. Exact match for the 5 anchor. | 5 / 5 |
Workflow Clarity | Prerequisites are sequenced (install packages → configure RPC → set endpoint), each instruction is an unambiguous single command, and the Error Handling table pairs causes with solutions (e.g., "RPC Rate Limit → Switch to backup endpoint or wait"). There is no explicit setup-verification step before running simulations, which leaves a minor validation gap at anchor 4; the destructive/batch cap does not apply since the skill only simulates "without executing real transactions", and error-recovery guidance keeps it above anchor 3. | 4 / 5 |
Progressive Disclosure | The body is a lean overview with well-signaled, one-level-deep references — references/implementation.md "for detailed output examples and risk scoring methodology", references/errors.md "for comprehensive error handling", references/examples.md "for multi-provider comparison and backtesting examples" — all of which are real files, with implementation scripts correctly separated into scripts/. Minor blemishes (implementation.md is linked from both Output and Resources, and two Resources entries lack links) do not materially harm navigation, so it best fits the 5 anchor. | 5 / 5 |
Total | 17 / 20 Passed |