Content
96%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 knowledge skill: it documents only what Claude cannot infer about this codebase, anchors every rule to named functions and real failure stories, and closes with executable validation. The only structural note is that everything lives inline in one file rather than being split across references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with codebase-specific, non-inferable invariants (the overlay base-vs-persistent split, owningPluginUri identity keys, the const-enum type-only-import crash, the test runner not failing on type errors) and contains no explanation of concepts Claude already knows; every section carries unique information with no padding. | 5 / 5 |
Actionability | Guidance is fully concrete: named functions and files (resolveCustomizationEnablement(), targetForMcpServer(), createNoopCustomizationEnablementService()), copy-paste validation commands, a grep-able log signature ("tool_search override … clientMatched=[…]"), and a specific cross-session reproduction recipe. | 5 / 5 |
Workflow Clarity | Sections are sensibly ordered (files to read before changing, mental model, the routing rule, per-provider enforcement traps, debugging playbook, validation), with explicit validation commands, a reproduction recipe, and a known-false-alarms list providing feedback loops for error recovery. | 5 / 5 |
Progressive Disclosure | Well-organized single-file structure with clear section headers and clearly signaled cross-skill pointers (agent-host-logs, launch), but all 126 lines are inline with no content split into bundle files — the under-50-line exception does not apply, and per-provider enforcement minutiae or gotchas could plausibly live in a reference file. | 4 / 5 |
Total | 19 / 20 Passed |