Feedback on your mastering-aws-cdk skill #2
Description
I took a look at your mastering-aws-cdk skill and wanted to share some thoughts.
Links:
The TL;DR
You're at 91/100, solid A territory. This is based on Anthropic's skill best practices rubric. Your strongest area is the architecture—the way you've structured the skill (main SKILL.md at ~130 lines delegating to 7 reference files) is exactly how Progressive Disclosure should work. Weakest area is Spec Compliance (12/15), mostly because you're only hitting 1-2 trigger phrases when you could expand that.
What's Working Well
- Architecture is chef's kiss. You've nailed the layered structure—SKILL.md stays focused and delegates deep content to reference files. No nested rabbit holes. This is how token economy gets done right.
- Actual troubleshooting teeth. The debugging checklist with copy-pasteable checkboxes is the kind of thing that saves people 30 minutes at 2am. That's real utility.
- Service patterns are thorough. Lambda, DynamoDB, API Gateway sections cover the pain points people actually hit—not generic AWS docs regurgitation. The OIDC setup specifically? That's niche enough that it matters.
- Consistent terminology throughout. L1/L2/L3 constructs, synth/diff/deploy workflow—you're not mixing metaphors. Developers can search and find stuff.
The Big One: Beef Up Your Trigger Phrases
Right now you've got maybe 1-2 trigger phrases in the description ("cdk deploy", "cdk synth"). You're leaving discoverability on the table. Expand that description to cover actual queries people type:
Current: "AWS CDK production patterns and best practices"
Better: "AWS CDK L2 constructs, production deployment patterns, OIDC federation, cdk-nag security scanning, troubleshooting CDK errors, infrastructure-as-code patterns, cdk synth, cdk diff, cdk deploy, CDK bootstrap"
This bump alone gets you points in Ease of Use (discoverability criterion) without changing any actual content. Just making the skill findable.
Other Things Worth Fixing
-
Minor redundancy across service sections. Lambda and DynamoDB both say "set log retention explicitly" in slightly different ways. Pull that into one canonical observability section and reference it. Small consolidation, tighter writing.
-
Workflow step formatting. You number them (1) Intake, (2) Stack Design—great start. But the sub-bullets under each could use explicit numbering too. Makes it scannable. Goes in SKILL.md around line 45.
-
Add a version field to frontmatter. Just
version: 1.0.0in your YAML block. Helps with tracking updates, follows convention. One line, zero friction.
Quick Wins
- Expand trigger phrases in description (biggest bang for buck—gets you +2-3 points)
- Consolidate redundant best practices into observability reference
- Add version to frontmatter
- Format sub-steps with explicit numbering for better scannability
Checkout your skill here: SkillzWave.ai | SpillWave We have an agentic skill installer that install skills in 14+ coding agent platforms. Check out this guide on how to improve your agentic skills.