Content
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is appropriately concise and well-structured for a simple single-purpose skill, with a clear sequence and a re-run feedback loop. Its main weakness is actionability: it tells Claude what to do but not how, with no executable commands, tool examples, or output format.
Suggestions
Add concrete guidance for discovering the repo's build/type-check commands (e.g., read package.json scripts, Makefile targets, or CI config) and show an example invocation.
Specify the expected error-summary format with a short example output so the "Summarize errors by file and type" step is unambiguous.
Make the feedback loop explicit: define what "blocked" means and instruct Claude to report remaining blockers with the failing command output when it cannot proceed.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean with no padding or explanations of concepts Claude already knows ("Compile or type-check failures are blocking local validation or CI" followed by a terse numbered workflow), matching the level-3 anchor "Lean and efficient; assumes Claude's competence; every token earns its place". | 3 / 3 |
Actionability | Steps like "Run the repo's compile and type-check commands" and "Summarize errors by file and type" give direction but no executable commands, tool names, or example output format, fitting the level-2 anchor "Some concrete guidance but incomplete... missing key details" rather than copy-paste-ready level 3. | 2 / 3 |
Workflow Clarity | The four steps are clearly sequenced and include a re-run loop ("Re-run checks until clean or blocked"), but the validation checkpoint and error-recovery handling are implicit (no explicit handling of what "blocked" means or how to surface remaining blockers), matching the level-2 anchor "sequence present but checkpoints missing or implicit". | 2 / 3 |
Progressive Disclosure | The skill is under 50 lines, needs no external references, and is organized into clearly labeled Trigger/Workflow/Output sections, so per the rubric's simple-skills guidance it qualifies for the level-3 anchor "well-organized sections" without separate files. | 3 / 3 |
Total | 10 / 12 Passed |