opencode-cli

I took a look at your opencode_cli skill and wanted to share some thoughts.

Links:

• GitHub
• SkillzWave marketplace

The TL;DR

You're at 84/100, solidly in B territory. This is based on Anthropic's skill best practices rubric. Your strongest area is Progressive Disclosure Architecture (27/30) — the skill is well-structured with focused reference files that don't bloat the main file. Weakest spot is Utility (17/20) — the skill solves a real problem, but the feedback loops and error handling could be tighter.

What's Working Well

• Progressive Disclosure is chef's kiss. You've got 6 focused reference files that delegate responsibility perfectly. SKILL.md stays lean at 195 lines while each reference goes exactly one level deep. That's the right structure.
• Discovery-friendly. Your trigger phrases are solid — "configure opencode", "vertex ai", "mcp servers", "ollama", "local models" — these will actually get Claude to activate the skill when people ask about that stuff.
• Token economy. Minimal redundancy in the main file. You're not repeating yourself across sections, which keeps things efficient.
• Practical examples. The JSON config templates and Python subprocess integration patterns give people real, copy-paste-ready code to work with.

The Big One: Reference Files Need Navigation

Your reference files are 200-380 lines but they're missing tables of contents. cloud-providers.md (351 lines), integration-patterns.md (217 lines) — when someone lands in these files looking for a specific provider or pattern, they're scrolling blind.

The fix: Add a ## Table of Contents section at the top of each reference file with anchor links. Takes 10 minutes, instantly improves usability, and probably gets you +1-2 points toward Ease of Use.

## Table of Contents
- [Anthropic](#anthropic)
- [Google Vertex AI](#google-vertex-ai)
- [OpenAI](#openai)

Other Things Worth Fixing

1. Duplication in SKILL.md (line 122-124). Your Vertex AI section just says "see vertex-ai-setup.md" — but that's already in your Reference Guides table. Redundant redirect. Delete it and save tokens.

2. Second-person voice slips. Lines like "you can" creep into integration-patterns.md. Stick to imperative: "Configure the model" not "You can configure the model." Small thing, but consistency matters.

3. Numbered steps in local-models.md. The Ollama setup is just prose blocks — no clear 1-2-3 workflow. Users want: "1. Install Ollama 2. Start server 3. Pull models 4. Configure OpenCode." That's +2 points right there.

4. Workflow clarity in references. Your main file has "Quick Start" (4 numbered steps), but reference files lack that structure. Add numbered workflows to provider-config.md and mcp-servers.md.

Quick Wins

• Add TOCs to files over 100 lines (+1-2 points)
• Remove redundant Vertex AI redirect in SKILL.md
• Add numbered steps to Ollama setup
• Fix "you can" voice consistency
• Structure cloud-providers.md with sequential steps

Get these done and you're looking at 87-89/100 — A-territory.

Checkout your skill here: SkillzWave.ai | SpillWave We have an agentic skill installer that installs skills in 14+ coding agent platforms. Check out this guide on how to improve your agentic skills.