Content
50%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.
The skill is rich in executable commands and has a sensible four-step workflow with a validation step, but it is substantially overweight: it duplicates its own frontmatter and its reference files, and leaves the scripts bundle unlinked. Destructive bulk examples lack the validation checkpoints the skill's own principles require.
Suggestions
Cut the body to a true overview: drop the duplicated When-to-Use/trigger-phrase/service-list sections (they live in the description) and move the per-service 'Power Commands', 'Examples', and 'Best Practices' sections into the existing reference files, keeping only one representative snippet per service inline.
Fix or remove the broken snippets: the custom-polling example (no loop, uninitialized i) and the DynamoDB query's expression-attribute-values ("{"=":...}" should be ":pk"), and add a --dry-run/confirmation checkpoint before the bulk stop-instances and delete-log-stream examples.
Link the scripts/ bundle from the body — e.g. a 'Shell aliases: see scripts/aws-blast.sh' and 'JMESPath templates: see scripts/jmespath-templates.json' section — so the bundled tooling is discoverable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~440-line body re-derives the frontmatter (9 'When to Use' bullets, 10 trigger phrases, a duplicate service-coverage list) and then inlines ~300 lines of per-service command catalogs, examples, and best practices that duplicate the four 400-line reference files. This is systematic padding, not incidental over-explanation. | 2 / 5 |
Actionability | Mostly executable, copy-paste-ready AWS commands covering the common cases (waiters, presign, IAM audit loop, bulk stop). Minor gaps: the 'custom polling with exponential backoff' snippet is broken shell (no loop, uninitialized i), and the DynamoDB query uses an invalid expression-attribute-value key ('=' instead of ':pk'). | 4 / 5 |
Workflow Clarity | Steps 1-4 are clearly sequenced with an explicit 'Validate and Verify' step, but the destructive/batch example workflows (piping describe-instances straight into stop-instances, xargs-deleting log streams) skip the dry-run/confirm checkpoints the skill itself mandates. Per the rubric, batch/destructive workflows without integrated validation cap at 3. | 3 / 5 |
Progressive Disclosure | The four references/ files are real and listed with one-line scopes in Step 2, but the body re-inlines their per-service material instead of remaining an overview, and the scripts/ bundle (aws-blast.sh, jmespath-templates.json) is never referenced anywhere, making it undiscoverable. | 3 / 5 |
Total | 12 / 20 Passed |