Content
65%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.
The body is a concise, well-organized rule set for a meta-prompting research workflow with a genuinely lean token footprint. Its main weaknesses are undefined prerequisites (Rosetta prep steps, load-context skill, FEATURE TEMP folder) and the absence of a sequenced workflow or meta-prompt template, which leave actionability and workflow clarity at the midpoint. Adding a short numbered sequence and a skeleton of the generated research prompt would make it fully executable.
Suggestions
Replace the flat rule lists with a numbered workflow (prep → craft meta-prompt → spawn subagent → track state → self-validate) so the sequence and checkpoints are explicit.
Define or link the undefined prerequisites: what "Rosetta prep steps" and the "load-context skill" are, and what/where the "FEATURE TEMP folder" is.
Include a compact template or example of the generated research prompt so the skill's central artifact is concrete rather than described.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean rule-list with no explanations of concepts Claude already knows, assuming competence throughout. Minor filler remains ("MUST prioritize ACCURACY over SPEED", "Use synthesis and comparison approach") that could be trimmed, matching anchor 4 rather than 5's every-token-earns-its-place. | 4 / 5 |
Actionability | Concrete guidance exists ("research-state.md" in FEATURE TEMP folder, "docs/<feature>-research.md", DeepWiki, Context7, todo tasks), but key details are missing: "Rosetta prep steps", "load-context skill", and "FEATURE TEMP folder" are undefined, and there is no template or example of the meta-prompt that is the skill's core artifact. This fits anchor 3 (some concrete guidance but incomplete, missing key details) rather than 4's minor gaps. | 3 / 5 |
Workflow Clarity | The rules imply a sequence (plan → incremental tasks → update state → self-validate) and an explicit final self-validation checkpoint exists ("Always create self-validation task at the end to re-review all conclusions"), but steps are jumbled across two flat rule lists with no ordered numbering and intermediate checkpoints only implicit. Anchor 3 (sequence present, checkpoints missing or implicit) fits better than 4's clear sequence with most checkpoints. | 3 / 5 |
Progressive Disclosure | There are no bundle files and the body is roughly 50 lines of single-task content organized into clear role/when-to-use/core-concepts/process sections. Per the rubric's simple-skill exception (under 50 lines, no need for external references), well-organized sections alone warrant a 5. | 5 / 5 |
Total | 15 / 20 Passed |