Content
75%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 well-crafted, instruction-only skill body: modes are precisely specified with an invocation table, a complete source_type mapping, concrete search commands, and exact output templates, and it includes a validation pass plus logging. The main gaps are pseudocode for the core map-building step, a validation step that is conditional on an optional skill, and no progressive disclosure into reference files despite four sizable output templates.
Suggestions
Replace the pseudocode source-map construction with a concrete, executable command or short script (e.g., a jq/python one-liner over .manifest.json) so Step 1 is copy-paste runnable like the rest of the skill.
Add a non-conditional self-check after building the source map (e.g., verify every mapped page exists in index.md and counts are nonzero) so validation does not depend on the optional impl-validator skill being available.
Move the four mode-specific output templates (Browse, Diff, Map, log line) into a references/ file and keep a one-line pointer per mode in SKILL.md to slim the main body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: a compact invocation-to-mode table, a terse source_type mapping, and output templates rather than explanations of what a wiki or manifest is. A few asides could still be trimmed ('these are the richest nodes', 'A page can appear in multiple tools' sets'), placing it at anchor 4 rather than the fully-lean 5. | 4 / 5 |
Actionability | Mostly executable guidance: concrete commands ('rg -l "<topic>"', grep for '^(title|category|tags|updated):'), an exact source_type mapping table, and copy-ready output templates for every mode. Not a 5 because the central map construction is pseudocode ('tool_pages = { ... set(...) }') with no concrete command for parsing .manifest.json; not a 3 since the rest is directly runnable. | 4 / 5 |
Workflow Clarity | Clear sequence: resolve config -> read manifest/index -> build source map -> execute mode -> validate -> log, with an explicit validation step (impl-validator with a concrete checks list) and a feedback loop ('Apply any issues it surfaces before presenting output'), plus error handling for a missing manifest. Not a 5 because validation is conditional ('if available') and does not cover the map-building step itself; not a 3 because checkpoints are explicit and operations are read-only, so the destructive-batch cap does not apply. | 4 / 5 |
Progressive Disclosure | No bundle files exist, and the single-file structure is well organized with clear section headers and one clearly-signaled external reference (the Config Resolution Protocol in llm-wiki/SKILL.md, one level deep). Not a 5 because at ~150 lines with four mode-specific output templates, some content could be split into reference files; not a 3 because nothing is buried and the structure is easy to navigate. | 4 / 5 |
Total | 16 / 20 Passed |