Content
36%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 functions as a design/specification document for the skill rather than an operational guide: it catalogs capabilities, algorithms, and output formats, but provides no executable steps, commands, or code Claude could actually run, and mixes in project-management material (roadmap, contributing phases, PRD references). Structure is decent, but detail is inlined that belongs in reference files, and the one cross-reference points to a nonexistent file.
Suggestions
Replace the descriptive pipeline sections with a sequenced, executable workflow (detect inputs → extract tokens → analyze components → generate outputs) with explicit validation checkpoints tied to the quality gates, plus fix-and-retry guidance for the listed error codes.
Move implementation-status, contributing/roadmap, and research-sources material out of SKILL.md — none of it helps Claude execute the skill — and drop the time-sensitive "(2026)" reference.
Split the output-structure tree, error-code catalog, and architecture/API heuristics into one-level-deep reference files under references/ (with clear links from SKILL.md), and either ship `skills/blocks/codex-host-adapter.md` or remove the dangling reference to it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly dense, structured material (tables, lists, pipelines) rather than explanations of known concepts, but it is padded with developer-project content that earns no tokens at runtime — "Implementation Status" checklists, week-by-week "Contributing" phases, "Research Sources", "Modern reverse-engineering practices (2026)" (time-sensitive), and a PRD reference. This fits anchor 3 (mostly efficient with some unnecessary sections) rather than 4, where only minor trimming would be needed. | 3 / 5 |
Actionability | The body describes what the skill does ("AST parsing for TypeScript/JavaScript", "Uses CIEDE2000", K-means clustering, service-boundary heuristics) but never instructs Claude how to execute any of it — the only commands are invocations of the skill itself (`/octo:extract ./my-app --mode design`), and no runnable code or concrete steps appear. This matches anchor 2 (high-level hints, missing the specific steps to execute); it is not a 3 because there is no even partially executable guidance for performing an extraction. | 2 / 5 |
Workflow Clarity | No extraction workflow is sequenced: there is no step order, no validation checkpoints wired into a process, and no fix-and-retry loop — error codes (`ERR-001`..`VAL-004`) and quality gates are cataloged but no handling guidance connects them. Quality gates do exist as a validation concept, so this sits between anchor 1 (steps missing, no validation) and anchor 3 (steps listed with checkpoint gaps) — anchor 2's "rough sequence... validation absent" fits best since the usage examples imply an order but define no operational steps. | 2 / 5 |
Progressive Disclosure | The single SKILL.md is well-sectioned but monolithic: content that clearly belongs in separate reference files (the 40-line output-directory tree, error-code catalog, architecture heuristics, performance tables) is inlined, and no references/, scripts/, or assets/ directories exist. The one referenced file, `skills/blocks/codex-host-adapter.md`, is not present in the bundle. Anchor 3 (some structure, content that should be separate is inline) fits; it is not a 2 because section headers make it navigable. | 3 / 5 |
Total | 10 / 20 Passed |