Content
100%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.
An exemplary instruction-only skill body: lean, highly actionable within the justified constraint of an un-provisioned private-beta API, with a clearly sequenced workflow, validation checkpoints, and well-signaled one-level-deep references that all exist on disk.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean with no concept-explanation padding (no 'what is a DON' or 'what is EIP-712'); every section (routing table, working rules, safety invariants, freshness, feedback loop) earns its place and assumes Claude's competence. | 5 / 5 |
Actionability | Although code-free (justified because the private-beta client API is not publicly provisioned), it gives concrete actionable guidance: explicit fail-closed conditions, 'use finalized for irreversible business actions', durable idempotency by immutable event ID, and an explicit pre-submission confirmation checklist (chain selector, Smart Account, operation ID, deadline, ordered tx with to/value/intent, signer identity, atomic failure effect). | 5 / 5 |
Workflow Clarity | The workflow is clearly sequenced (classify request -> route by artifact fit -> establish provisioning -> request official docs or provide labeled pseudocode -> preserve user values -> explicit approval before submission) with validation checkpoints (event verification, fail-closed, approval gate) and a feedback-loop section; destructive/batch operations carry explicit validation (atomic Operations, approval before irreversible action). | 5 / 5 |
Progressive Disclosure | SKILL.md is a clear overview/routing layer pointing to four verified one-level-deep references (concepts.md, events.md, operations.md, dta.md) with an explicit numbered load order, 'Load only the reference needed', and 'Do not load references speculatively'. | 5 / 5 |
Total | 20 / 20 Passed |