Content
82%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 dense, high-signal interrogation script: six questions with concrete thresholds, an executable workflow, a copy-paste output template, and routing. Its weaknesses are operational: the workflow's three commands reference scripts absent from the bundle with undefined JSON input shapes, and the Related section's cross-references all resolve to nothing.
Suggestions
Bundle the three referenced Python scripts (model_buildvsbuy_calculator.py, ai_risk_classifier.py, ai_cost_economics.py) under the skill's own scripts/ directory, or inline their decision logic, so the Workflow commands are runnable from this skill's location.
Define the expected shape of use_case.json and workload.json (or point to where they are defined) so the workflow commands are copy-paste executable rather than illustrative.
State how script outputs feed the six questions — e.g., a checkpoint after risk classification ('if PROHIBITED, stop and re-scope before proceeding to cost economics') — to close the sequence gap between the questions and the Workflow block.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every section carries project-specific judgment Claude cannot infer — '50-100 representative inputs minimum', 'breakeven typically 1-10B tokens/month for 70B-class', the AI engineer vs ML engineer vs research scientist distinctions — with no filler explaining known concepts. Lean and efficient. | 5 / 5 |
Actionability | The Workflow block gives three concrete python commands and the Output Format template is copy-paste ready, but the commands point to '../../../c-level-advisor/skills/chief-ai-officer-advisor/scripts/*.py' which do not exist in this bundle, and the use_case.json / workload.json input shapes are never defined. Mostly executable with gaps, not fully executable as shipped. | 4 / 5 |
Workflow Clarity | A numbered 3-step script sequence, per-question decision rules ('PROHIBITED → cannot launch in EU; re-scope'), and a verdict gate (🟢/🟡/🔴) give clear sequence with most checkpoints present. Minor gaps: no stated ordering between the six questions and the three scripts, and no re-check loop after script output. | 4 / 5 |
Progressive Disclosure | A well-sectioned single file at reasonable length with clearly labeled Related references one level deep. However, every referenced external path (the advisor skill scripts, ../../agents/cs-caio-advisor.md, the adjacent chief-data-officer skill) is broken in this bundle, and no scripts/ or references/ directories ship with the skill — minor organization gaps. | 4 / 5 |
Total | 17 / 20 Passed |