Content
77%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 body is a highly actionable, well-sequenced runbook with concrete commands, explicit validation, and error-recovery loops. Its main weaknesses are repeated rules that pad the token budget and inlined reference-grade tables that would benefit from separate files now that no bundle exists.
Suggestions
De-duplicate rules restated across sections — consolidate the "never spawn npm run dev/vite" and "never guess the port" guidance into a single Hard Rules section and reference it elsewhere rather than repeating.
Move the error-code table and state-files table into a references/ file (e.g. references/cli-reference.md) and keep a short summary inline, improving progressive disclosure and token efficiency.
Trim narrative filler such as "Inspired by Codex Sites' saved version model:" so every remaining sentence carries operational weight.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient operational detail Claude would not know, but several rules are restated across sections (e.g. "Never spawn npm run dev/vite" and "NEVER guess the port" each appear twice) and a few lines are filler ("Inspired by Codex Sites' saved version model"). | 3 / 5 |
Actionability | Provides copy-paste-ready CLI commands with exact flags and full MCP tool calls with parameters (e.g. manageApps deployApp with serviceName/filePath/buildPath/framework values), plus a concrete CLI-path resolution sequence covering the common cases. | 5 / 5 |
Workflow Clarity | The save→deploy→record→rollback flow is explicitly sequenced with validation checkpoints (preview --status, parse JSON stdout) and error-recovery feedback loops (envQuery retry on env errors, error-code table with recovery actions), including for the destructive rollback. | 5 / 5 |
Progressive Disclosure | Well-sectioned with clear headers and clearly signaled on-demand references (knowledge-base skills via searchKnowledgeBase), but no bundle files exist and bulk content like the error-code and state-file tables is inlined rather than split into reference files. | 3 / 5 |
Total | 16 / 20 Passed |