Content
50%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 organized with sensible sections and a coherent assess-plan-apply-test flow, but every dimension stops short of concrete: no specific techniques or examples are provided, verification checkpoints are implicit, padding sections consume tokens, and the playbook that would supply the detail is referenced but missing from the bundle. It reads as a solid skeleton awaiting its substance.
Suggestions
Fix the dangling reference: either add `resources/implementation-playbook.md` with the detailed patterns and examples the body promises, or remove the two pointers to it in the Instructions and Resources sections.
Tighten token use by deleting the persona opener ("You are a code refactoring expert...") and the "Context" and "Requirements" sections, which restate the description and the user's request without adding actionable content.
Make verification an explicit feedback loop in the workflow, e.g. "After each slice, run the test suite; if it fails, revert the slice before proceeding" and "Only proceed to the next step when tests pass", replacing the implicit "keep behavior stable" goal.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is short and scannable, but three sections add little: the persona opener "You are a code refactoring expert specializing in...", the "Context" section that restates the description ("The user needs help refactoring code to make it cleaner, more maintainable..."), and the circular "Requirements" section ("Take the requirements, target, and scope from the user's request"). This is anchor 3 (mostly efficient but could be tightened): below anchor 4 because the padding spans multiple sections rather than minor instances, above anchor 2 because no known concepts are explained and the core is lean. | 3 / 5 |
Actionability | Instructions give a real ordered method ("Assess code smells, dependencies, and risky hotspots", "Propose a refactor plan with incremental steps", "Apply changes in small slices and keep behavior stable", "Update tests and verify regressions") but without specific techniques, commands, or examples, and the file that would carry the detail ("open `resources/implementation-playbook.md`") does not exist in the bundle. This is anchor 3 (some concrete guidance but incomplete, missing key details): above anchor 2 because the bullets are executable directives rather than vague hints, below anchor 4 because the specifics needed to execute are deferred to a missing file. | 3 / 5 |
Workflow Clarity | The Instructions are clearly sequenced (assess, plan, apply in slices, update tests) and Safety adds "ensure tests pass" and "Avoid changing external behavior without explicit approval", but checkpoints are implicit goals ("keep behavior stable") rather than explicit gates, and there is no feedback loop for what to do when tests fail during a behavior-sensitive refactoring. This matches anchor 3 (sequence present but checkpoints missing or implicit): above anchor 2 because the sequence and a verification step exist, below anchor 4 because validation is not tied to explicit per-step checkpoints. | 3 / 5 |
Progressive Disclosure | Sections are well organized with clear use/not-use scoping and the reference is clearly signaled ("If detailed patterns are required, open `resources/implementation-playbook.md`"), but no bundle directories exist and `resources/implementation-playbook.md` is absent, so both references (Instructions and Resources) are dangling and navigation breaks exactly where disclosure should happen. This lands at anchor 3: the structure is good, but a broken reference path is a worse defect than anchor 4's "minor organization gaps", and the simple-skill exception does not apply since the body explicitly depends on the external file. | 3 / 5 |
Total | 12 / 20 Passed |