Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a strong, highly actionable deployment skill with excellent workflow clarity and a well-thought-out decision tree covering multiple environments and states. Its main weakness is length — the comprehensive coverage of all deployment scenarios makes it verbose, and the inline treatment of five distinct methods plus three agent-specific variants could benefit from better progressive disclosure. The skill demonstrates deep domain knowledge of Vercel CLI behavior and edge cases.
Suggestions
Consider extracting the five deployment methods into a separate reference file (e.g., METHODS.md) and keeping only a decision-tree summary in SKILL.md to reduce token cost
Consolidate repeated patterns — the team selection logic and 'vercel deploy -y --no-wait --scope <team-slug>' pattern appears in multiple methods and could be stated once with back-references
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is quite long (~200+ lines) and includes some redundant explanations (e.g., re-explaining team selection logic multiple times across methods, explaining what .vercel/project.json vs .vercel/repo.json contain). However, most content is genuinely necessary given the complex decision tree. Some tightening is possible — the `.vercel/` directory explanation block and repeated deploy patterns could be consolidated. | 2 / 3 |
Actionability | Excellent actionability throughout. Every method includes specific, executable bash commands with correct flags. The skill provides concrete CLI invocations (vercel deploy -y --no-wait, vercel link --repo --scope, vercel inspect), exact script paths, and specific examples for each deployment scenario. Copy-paste ready. | 3 / 3 |
Workflow Clarity | The workflow is exceptionally well-structured: Step 1 gathers state with four specific checks, Step 2 branches into clearly defined methods based on state combinations (linked+git, linked+no-git, unlinked+auth, unlinked+no-auth, fallbacks). Each method has numbered steps with explicit validation points (e.g., 'Ask the user before pushing', 'vercel inspect to check status'). Feedback loops are present for auth failures (fall back to no-auth script). | 3 / 3 |
Progressive Disclosure | The skill references external scripts (deploy.sh, deploy-codex.sh) appropriately, but the main SKILL.md itself is quite long and monolithic. The agent-specific notes, troubleshooting, and output sections are well-separated, but the core Step 2 section contains five full deployment methods inline that could benefit from being split into separate files. No bundle files were provided to verify referenced paths. | 2 / 3 |
Total | 10 / 12 Passed |