I was curious how you'd approach cost estimation for LLM applications—there's a lot of variables to juggle. Your skill handles the complexity pretty cleanly, though the Progressive Disclosure could be tightened up to save tokens in the common path.
Links:
The TL;DR
You're at 81/100, solid B territory. This is based on Anthropic's skill best practices. Your strongest area is Writing Style (9/10)—clean, objective, no marketing fluff. The weakest link is Progressive Disclosure Architecture (18/30)—you've got good content, but it's all crammed into one 438-line file without the supporting structure that saves tokens and improves navigation.
What's Working Well
- Metadata is locked in — Your description nails the trigger phrases (planning budgets, build vs buy, TCO analysis), so users will find this when they need it
- Practical templates — The Project Budget Template and Quick Estimate sections give people a real starting point, not just theory
- Solid pricing foundations — Including concrete AWS/GCP pricing tables makes this immediately useful for actual estimation work
- Consistency — TCO, infrastructure costs, and development cost frameworks stay consistent throughout; no terminology confusion
The Big One: Missing Reference Files
Your references section mentions cloud-pricing.md and build-vs-buy.md (line 436-437), but these files don't exist. This is a -5 point hit because it breaks the Progressive Disclosure chain—you're promising layered detail that never materializes.
Fix: Either create those two files and move your detailed pricing tables (currently 77 lines, lines 44-121) into cloud-pricing.md, or strip the references section. If you go the file route, keep a summary table in SKILL.md and link deeper: "See Cloud Pricing Reference for complete EC2, GPU, and database pricing tables." This cuts token bloat for the common path while keeping detail available.
Impact: +5 points
Other Things Worth Fixing
-
Add a Table of Contents — At 438 lines, you need a TOC after the frontmatter (this is a >100 line requirement). Users shouldn't have to scroll to find "Infrastructure Cost Reference" or "Development Cost Estimation." Impact: +2 points
-
No procedural steps — You provide templates and data, but no numbered workflow. Add a "Cost Estimation Workflow" section: 1) Define scope, 2) Identify cost categories, 3) Calculate infrastructure costs from tables, 4) Apply development framework, 5) Validate against benchmarks. Users shouldn't have to infer the sequence. Impact: +3 points
-
Missing validation loops — Your Build vs Buy framework compares options, but there's no "how do I know this estimate is reasonable?" section. Add sanity checks: infrastructure should be 15-25% of total TCO, team costs include 30% overhead, compare against your own benchmarks. Impact: +1 point
Quick Wins
- Create
references/cloud-pricing.md and move detailed tables (biggest token win)
- Add TOC right after frontmatter
- Insert numbered workflow steps before templates
- Add 3-4 validation checks to catch bad estimates
- Keep the writing and templates—those are solid
Checkout your skill here: [SkillzWave.ai](https://skillzwave.ai) | [SpillWave](https://spillwave.com) 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.
I was curious how you'd approach cost estimation for LLM applications—there's a lot of variables to juggle. Your skill handles the complexity pretty cleanly, though the Progressive Disclosure could be tightened up to save tokens in the common path.
Links:
The TL;DR
You're at 81/100, solid B territory. This is based on Anthropic's skill best practices. Your strongest area is Writing Style (9/10)—clean, objective, no marketing fluff. The weakest link is Progressive Disclosure Architecture (18/30)—you've got good content, but it's all crammed into one 438-line file without the supporting structure that saves tokens and improves navigation.
What's Working Well
The Big One: Missing Reference Files
Your references section mentions
cloud-pricing.mdandbuild-vs-buy.md(line 436-437), but these files don't exist. This is a -5 point hit because it breaks the Progressive Disclosure chain—you're promising layered detail that never materializes.Fix: Either create those two files and move your detailed pricing tables (currently 77 lines, lines 44-121) into
cloud-pricing.md, or strip the references section. If you go the file route, keep a summary table in SKILL.md and link deeper: "See Cloud Pricing Reference for complete EC2, GPU, and database pricing tables." This cuts token bloat for the common path while keeping detail available.Impact: +5 points
Other Things Worth Fixing
Add a Table of Contents — At 438 lines, you need a TOC after the frontmatter (this is a >100 line requirement). Users shouldn't have to scroll to find "Infrastructure Cost Reference" or "Development Cost Estimation." Impact: +2 points
No procedural steps — You provide templates and data, but no numbered workflow. Add a "Cost Estimation Workflow" section: 1) Define scope, 2) Identify cost categories, 3) Calculate infrastructure costs from tables, 4) Apply development framework, 5) Validate against benchmarks. Users shouldn't have to infer the sequence. Impact: +3 points
Missing validation loops — Your Build vs Buy framework compares options, but there's no "how do I know this estimate is reasonable?" section. Add sanity checks: infrastructure should be 15-25% of total TCO, team costs include 30% overhead, compare against your own benchmarks. Impact: +1 point
Quick Wins
references/cloud-pricing.mdand move detailed tables (biggest token win)Checkout your skill here: [SkillzWave.ai](https://skillzwave.ai) | [SpillWave](https://spillwave.com) 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.