Content
92%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.
A dense, opinionated, highly actionable guide that assumes Claude's competence and pushes detail into one-level-deep reference files. Workflows — especially persisted agents — include explicit validation and feedback loops; the only slack is minor cross-section repetition.
Suggestions
State the '$fromAI cannot carry binary' rule once (it appears in the sub-node pattern, $fromAI, and binary sections) and cross-reference rather than repeating.
Condense the 'Integration with other skills' list to the most-referenced skills, noting the rest are already surfaced via the reference table.
Ensure each Anti-patterns table row adds a distinct, scannable fix rather than duplicating the narrative guidance above it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Information-dense and assumes Claude's competence throughout, but a few points are repeated across sections (the $fromAI-cannot-carry-binary rule, anti-patterns restated in the table) and the integration list is long — minor trim opportunities keep it just below fully lean. | 4 / 5 |
Actionability | Provides copy-paste-ready JSON connection snippets, exact `$fromAI()` expression syntax, concrete option names (maxIterations, autoFix, schemaType, contextWindowLength, passthroughBinaryImages), and a precise numbered MCP action sequence for persisted agents. | 5 / 5 |
Workflow Clarity | The persisted-agent build sequence is explicitly numbered with a validation step ('validate' before 'call' or 'publish'), a stale-hash feedback loop ('STALE_CONFIG — re-get and retry'), and a shipping checklist; the chat shell+core+sub-agents topology and quick-reference checklist reinforce it. | 5 / 5 |
Progressive Disclosure | Clear overview body with one-level-deep references signaled inline ('→ TOOLS.md') and summarized in a 'Reference files' table with 'Read when' guidance; cross-cutting concerns are delegated to named sibling skills rather than re-derived inline. | 5 / 5 |
Total | 19 / 20 Passed |