Content
28%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 opens well (use/don't-use triggers, brief instructions, safety notes) but then devolves into a padded persona resume enumerating technologies and principles Claude already knows. It provides almost no executable review procedure, output format, or validation guidance.
Suggestions
Delete the Expert Purpose, Capabilities, Behavioral Traits, and Knowledge Base sections — they enumerate standard knowledge (SOLID, Kafka, C4, DDD) Claude already has and consume most of the token budget; move any genuinely non-obvious criteria to a reference file.
Merge the Instructions and Response Approach lists into one sequenced review workflow with concrete checkpoints: what artifacts to request, what criteria to evaluate against, and explicit validation (e.g. confirm assumptions with the user before rating High/Medium/Low impact).
Add a concrete output format, such as findings ranked by architectural impact with trade-offs and a minimal ADR template, and replace the abstract Example Interactions with one worked example of input and expected review output.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly 130 of 170 lines are Capabilities, Behavioral Traits, and Knowledge Base sections that name-drop concepts Claude already knows (SOLID, Kafka, circuit breakers, the C4 model), plus two redundant step lists (Instructions and Response Approach); the efficient front sections keep it above anchor 1's 'severely verbose' but it is noticeably padded. | 2 / 5 |
Actionability | The guidance is abstract — "Gather system context", "Evaluate architecture decisions and identify risks", "Recommend improvements with tradeoffs" — with no concrete commands, output formats, review checklists, or worked examples, matching anchor 2 ('high-level hints but missing the specific steps to execute'). | 2 / 5 |
Workflow Clarity | Two competing, loosely defined step lists exist (Instructions 1-4 and Response Approach 1-8) with no validation checkpoints or error-recovery loops, matching anchor 2 ('Rough sequence present but many gaps; steps poorly defined; validation absent'). | 2 / 5 |
Progressive Disclosure | Sections are clearly labeled (no bundle files exist, so there are no broken references), but ~130 lines of capability and knowledge lists are inlined in SKILL.md that belong in reference files or should simply be cut, matching anchor 3 ('Some structure but could be better organized; content that should be separate is inline'). | 3 / 5 |
Total | 9 / 20 Passed |