Content
71%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 well-structured, highly actionable command workflow with genuine state-machine logic (existence checks, confirm-on-existing-doc, status routing), but it is a 745-line monolith that inlines three full document templates and repeats boilerplate, with zero use of bundle files. The stray duplicate frontmatter block at the top of the body adds further noise.
Suggestions
Move the three document templates (RESEARCH.md, IMPLEMENTATION.md, PROGRESS.md) into a `references/templates.md` file and link to them, cutting ~150 lines from SKILL.md.
Delete the stray duplicate frontmatter block at the top of the body and state the feature-name prompting rule once in Argument Parsing instead of repeating it in five subcommands.
Trim the "Important Guidelines" section — "Use AskUserQuestion Liberally", "Progress Tracking", and "Always Works Philosophy" largely restate guidance already embedded in the subcommand steps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bulk is tight, concrete instruction, but there is clear tightening room: ~150 lines of inline document templates, a trailing "Important Guidelines" section that re-explains basics Claude already knows ("Test changes as you make them", "Don't assume code works - verify it"), and the "Get Feature Name" prompting procedure repeated across five subcommands. It is not 2 because the padding is a minority of the file; it is not 4 because "Use AskUserQuestion Liberally" and "Always Works Philosophy" largely restate guidance already embedded in the subcommand steps. | 3 / 5 |
Actionability | Every subcommand gives exact file paths (`docs/{name}/RESEARCH.md`), fully specified AskUserQuestion payloads (question, header, multiSelect, options), and copy-paste-ready markdown templates for all three documents. There are no gaps — even the dynamic phase-selection options state their generation rule ("dynamically generated from phases found in IMPLEMENTATION.md, marking completed ones"), so it is not 4. | 5 / 5 |
Workflow Clarity | Each subcommand is a numbered sequence with explicit checkpoints: doc-existence verification ("Check if `docs/{name}/RESEARCH.md` already exists"), Overwrite/Append/Skip confirmation before destructive overwrites, and next-step chaining, with missing docs redirecting to the correct subcommand — plus the status subcommand handling every doc-combination state. It is not 5 because phase completion gates on generic "test as you go" advice rather than a defined, concrete verification checkpoint; it is not 3 because validation and redirect feedback loops are explicit throughout. | 4 / 5 |
Progressive Disclosure | This is a 745-line monolith with no references/, scripts/, or assets/ bundle at all; the three full document templates (~150 lines) and per-subcommand detail clearly belong in separate referenced files. It is not 2 because internal structure is strong (clear `## Subcommand:` and `### Step:` headers, explicit argument-routing table); it is not 4 because nothing is split out or referenced — everything is inlined in one file. | 3 / 5 |
Total | 15 / 20 Passed |