Content
42%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 well-structured with clear sectioning and a sensible workflow spine, but it is an instruction-only skill whose instructions rarely reach concreteness: the classification scheme it routes on is never defined, validation gates name tools without commands, and the Avoid and Anti-Patterns sections duplicate each other verbatim. Its only reference points outside the bundle while the actual reference files go unmentioned.
Suggestions
Define the classification criteria: give one-line hallmarks of existing-system, partial-system, and greenfield (e.g. what evidence in routes/components/conventions maps to each class).
Merge the duplicated Avoid and Anti-Patterns sections into one, and make validation steps concrete by naming the actual commands for 'Plugin Eval' and the 'strict skill audit'.
Fix progressive disclosure: replace the external archive pointer with references to the actual bundle files (references/contract.yaml, references/evals.yaml, references/task-profile.json) so the body signals where each kind of detail lives.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean and imperative, but the Avoid section is duplicated verbatim as Anti-Patterns (three identical bullets: 'Do not own token-governance work...', 'Do not own motion-only polish...', 'Do not ask a broad design interview...') and gate-stopping language repeats across Constraints and Validation, fitting the 3 anchor ('mostly efficient but includes some unnecessary explanation or could be tightened') rather than the 4 anchor, which requires only minor trims. | 3 / 5 |
Actionability | As an instruction-only skill the absence of code is acceptable per the scoring notes, but the guidance stays at high-level hints: 'Classify system maturity and implementation readiness' gives no criteria for the existing-system/partial-system/greenfield scheme, 'Resolve overlap against local frontend owner skills' offers no method, and 'Run Plugin Eval and strict skill audit' names no command — matching the 2 anchor ('minimal concrete guidance; high-level hints but missing the specific steps to execute') rather than 3, since even the central classification the skill routes on is undefined. | 2 / 5 |
Workflow Clarity | A clear six-step sequence is present with an ask-only-when-material rule and a Validation section demanding exact commands and pass/fail reporting, but the checkpoints are named abstractly ('Run Plugin Eval and strict skill audit') with no concrete commands or error-recovery loop, matching the 3 anchor ('sequence present but checkpoints missing or implicit') rather than 4. | 3 / 5 |
Progressive Disclosure | Sections are well-organized and the body is compact, but scored against the actual bundle: references/ contains contract.yaml, evals.yaml, and task-profile.json, none of which the body signals or links, and its single reference points outside the bundle ('Archived full context: Infrastructure/references/deferred-skill-context/agent-ops-frontend-design/'), matching the 3 anchor ('references present but not clearly signaled') rather than 4. | 3 / 5 |
Total | 11 / 20 Passed |