Content
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill functions as a routing meta-skill with reasonable structure and good safety guardrails (failure modes, anti-patterns, constraints). Its main weakness is that the core routing logic is entirely deferred to an unverifiable reference file, leaving the SKILL.md itself without enough actionable content to stand alone. The output schema is described but not exemplified with a concrete JSON object, and the inputs section is too vague to be useful.
Suggestions
Add a concrete JSON example of the routing handoff object output (showing all fields like schema_version, selected_lane, confidence, etc.) so Claude knows the exact shape to produce.
Include at least a summarized version of the routing decision logic from references/workflow.md inline, rather than deferring the entire workflow to an external file.
Make the 'Inputs' section more specific—define what 'request text' looks like, what 'constraints and trust requirements' means concretely, and provide an example input.
Expand the Examples section to show full input→output pairs including the complete handoff object, not just the lane name.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is moderately efficient but includes some unnecessary sections like 'Philosophy' (two bullet points that are somewhat generic), and the 'When to Use' section partially repeats the description. The 'Inputs' section is vague enough to be unhelpful. However, it avoids explaining concepts Claude already knows and stays relatively focused. | 2 / 3 |
Actionability | The skill provides some concrete guidance—example routings, a named output schema with specific fields, and a validation command—but the core routing logic is deferred entirely to 'references/workflow.md' which is not provided. The output object fields are listed but lack a concrete JSON example or schema definition. The examples section maps inputs to skill names but doesn't show the actual handoff object output. | 2 / 3 |
Workflow Clarity | The workflow is essentially 'use the detailed routing protocol in references/workflow.md' which defers the actual multi-step process. The failure mode and anti-patterns sections provide good guardrails, and there's a validation script, but the actual routing sequence is absent from this file. The 'stop at ambiguity' instruction is a good checkpoint, but the workflow itself is not clearly sequenced here. | 2 / 3 |
Progressive Disclosure | The skill references multiple external files (workflow.md, contract.yaml, evals.yaml, etc.) with clear signaling, which is good structure. However, no bundle files were provided to verify these references exist, and the core workflow being entirely in a reference file means this SKILL.md is somewhat hollow—it's an overview that depends heavily on references without providing enough standalone value. The deep relative paths (../../../../../) suggest potentially fragile navigation. | 2 / 3 |
Total | 8 / 12 Passed |