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.
A well-structured instruction skill with excellent progressive disclosure (all nine references real and mapped to output sections) and a clearly sequenced decision workflow with a time-order gate. The main weaknesses are redundancy — negative guidance repeated across three sections and a trigger list duplicating the description — plus a minor step-count mismatch with the workflow reference template and no worked output example.
Suggestions
Consolidate the three overlapping negative-guidance blocks (Core Function's 'should not' list, Hard Rules, and 'What This Skill Should Not Do') into the single Hard Rules section, which is the most forceful and best-placed home for them.
Delete or trim the 'Sample Triggers' section — its examples duplicate the trigger phrases already in the description and in 'Input Validation' examples.
Add one short worked example (e.g., a filled Section F table row showing variable → proposed role → recommended handling → rationale) and sync the 8-step Decision Logic with the 9-step references/workflow-step-template.md to remove the step-count mismatch.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is prescriptive rather than explanatory, but there is clear trimmable redundancy: negative guidance appears in three overlapping places (Core Function's "should not" list, Hard Rules, and the separate "What This Skill Should Not Do" section), and the "Sample Triggers" section largely duplicates the description's trigger phrases. Not 2 because there is no padded conceptual explanation — nearly every statement carries operational weight; not 4 because the duplicated sections and trigger list are unnecessary tokens. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is mostly executable: an 8-step decision logic with named reference modules per step, a mandatory output structure (Sections A–K) with specified table columns and handling categories ("must adjust / control", "do not adjust", "role uncertain"), and an exact out-of-scope redirect template. Minor gaps keep it below fully executable: no worked example of a filled output section, and some directives remain abstract ("identify the true confounding problem rather than list generic covariates"). | 4 / 5 |
Workflow Clarity | The 8-step sequence is clear and each step maps to a reference module, with a gating checkpoint ("Do not proceed with adjustment logic until time order is at least partially clarified"), input validation with an explicit redirect, and a role-uncertain escape hatch. It falls short of 5 because there is no explicit error-recovery loop, and the body's 8 steps are out of sync with the 9 steps in references/workflow-step-template.md, which the body directs the model to follow for step alignment. | 4 / 5 |
Progressive Disclosure | SKILL.md functions as an overview that explicitly routes each of the nine real, substantive, one-level-deep reference files to specific output sections (e.g., "references/adjustment-selection-rules.md → use when deciding which variables should be adjusted for... in Sections E and F"), and states the consequence of skipping them. All referenced paths exist and navigation is unambiguous. | 5 / 5 |
Total | 16 / 20 Passed |