Content
35%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's domain content (graph-type table, caption structure, audience templates, pitfall tables) is genuinely useful, but it is buried in template boilerplate and, critically, documents an API that does not exist: every Python and CLI example targets `scripts/graph_interpreter` while the actual bundle ships only a single-method `scripts/main.py`. The reference file is an empty stub, so the promised detailed guidance is not actually available.
Suggestions
Align all code and CLI examples with the actual bundle: either implement the documented `GraphInterpreter` API in `scripts/main.py` (or `scripts/graph_interpreter.py`) or rewrite the Quick Start, Core Capabilities, and CLI Usage sections around the real `GraphInterpretation.describe()` interface — remove every reference to the fabricated module, class, and methods.
Cut the template boilerplate: delete the three verbatim repetitions of the description (When to Use, Key Features, scope line), the generic Implementation Details section with its broken "See ## Workflow above" reference, and the best-practice explanations of things Claude already knows (correlation ≠ causation, report CIs) — this alone would roughly halve the body.
Move the graph-type table, audience explanation templates, Common Patterns, and Quality Checklist into `references/` (populating the currently empty `guidelines.md` or adding focused files), keeping SKILL.md to an overview plus a working quick start with clearly signaled one-level-deep links.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~400-line body repeats the frontmatter description verbatim three times ("Use this skill when the task needs Use when interpreting...", the "Key Features" restatement, and the scope line), pads with generic template text ("requires explicit assumptions, bounded scope, and a reproducible output format"), and explains concepts Claude already knows (correlation vs. causation, report CIs with point estimates, distinguish statistical from clinical significance). This is noticeably verbose with several padded sections, matching anchor 2 rather than 3 — the padding is not incidental but structural. | 2 / 5 |
Actionability | The core examples are not executable: the Quick Start imports `from scripts.graph_interpreter import GraphInterpreter` and demonstrates ~15 methods (`interpret`, `analyze`, `extract_statistics`, `generate_multi_audience`, `generate_caption`, `critical_appraisal`, `interpret_forest_plot`, `generate_grade_rating`, ...), and the CLI section invokes `python scripts/graph_interpreter.py`, but the only real script is `scripts/main.py` containing a `GraphInterpretation` class with a single `describe()` method — the documented API is fabricated. Only the Quick Check commands (`python -m py_compile scripts/main.py`, `python scripts/main.py`) actually run, leaving minimal genuinely concrete guidance (anchor 2). Not 1 because runnable verification commands and concrete output structures (caption template, statistical reporting example) do exist. | 2 / 5 |
Workflow Clarity | A sequenced 5-step Workflow exists with a fallback path ("If execution fails or inputs are incomplete, switch to the fallback path and state exactly what blocked full completion"), an Error Handling section, and pre-flight checkpoints ("stop early if the task would require unsupported assumptions"). However, the checkpoints are generic template language disconnected from the actual script's single-method capability, and "Implementation Details" opens with a broken forward reference ("See `## Workflow` above for related details") placed before the Workflow section. Steps are listed but validation is implicit and partly incoherent — anchor 3, not 4. | 3 / 5 |
Progressive Disclosure | The body does point to real bundle paths (`scripts/main.py` exists, `references/` exists), but it simultaneously references a nonexistent `scripts/graph_interpreter.py`, and `references/guidelines.md` is a 5-line heading-only stub while the body claims "Reference material available in `references/` for task-specific guidance" — references exist but do not clearly signal real content. Meanwhile ~300 lines of graph-type tables, audience templates, patterns, and checklists that belong in reference files are inlined in SKILL.md. Some structure, but organization and signaling are incomplete — anchor 3. | 3 / 5 |
Total | 10 / 20 Passed |