Content
85%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 well-structured hub-and-spoke skill: executable commands, BAD/GOOD examples, a workflow with real feedback loops, and a verified one-level-deep reference map. The only weaknesses are mild redundancy in the anti-pattern section and a GOOD example for shell safety that doesn't actually demonstrate a safe shell invocation.
Suggestions
In each anti-pattern, drop either the inline 'BAD: ... GOOD: ...' summary line or the code blocks — the duplicated pair doubles tokens without adding information.
Replace the shell-safety GOOD example with a genuine safe spawn (e.g., Bun.spawn(["grep", "-n", "pattern", fixedPath]) using an argument array without 'sh -c') so the fix matches the anti-pattern's own advice to use 'fixed command arguments'.
State the SQL prepared-statement requirement once — either in the workflow or in the Validate SQL Pattern command — instead of in both places plus the anti-pattern section.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — short command blocks with one-line expected outcomes, no concept tutorials — but each anti-pattern states its BAD/GOOD pair twice (an inline summary line followed by a code block restating it), and SQL validation guidance is repeated across the workflow and the Quick Commands section. Fits anchor 4 (minor trimming possible) rather than 5, and is clearly above anchor 3 since nothing explains concepts Claude already knows. | 4 / 5 |
Actionability | Commands are executable ("bun install", "rg -n \"query\\(|prepare\\(\" src") and BAD/GOOD code blocks are copy-paste ready, covering file I/O, SQL injection, and package management. Not a 5 because the shell anti-pattern's GOOD example shows a file read ("const safePath = Bun.file(userProvidedPath)") instead of a genuinely safe spawn with fixed argument arrays, leaving a minor gap for the common "run a command safely" case. | 4 / 5 |
Workflow Clarity | The 7-step workflow has explicit validation ("Validate behavior with tests or execution checks") and feedback loops for error recovery ("If bun install fails: check bun.lock/bun.lockb for merge conflicts... re-run", "revert the last change, confirm the error... before re-validating"), plus a hard gate on refactoring flagged raw SQL interpolations. Matches anchor 5; not a 4 because checkpoints are explicit rather than implicit. | 5 / 5 |
Progressive Disclosure | The body acts as a navigation hub with a "Reference Map" grouping all 13 bundle files by category, and every referenced path (e.g., references/runtime-globals.md, references/sqlite-basics.md) was verified to exist on disk, one level deep. Not a 4 because references are clearly signaled, categorized, and complete with no detail content inlined that belongs in them. | 5 / 5 |
Total | 18 / 20 Passed |