Content
65%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with executable code and clear decision routing, but it is large and monolithic: redundant 'Important Rules' restate earlier guidance, destructive/batch operations lack validation checkpoints, and substantial reference tables belong in separate files. Splitting reference material out and adding verification steps for risky operations would lift conciseness, workflow clarity, and progressive disclosure.
Suggestions
Move the enriched-data-field tables, the full signature-key table, and the GT/SAR method tables into separate reference files (e.g., REFERENCE.md, SIGNATURES.md) linked from a concise overview, to improve progressive disclosure and conciseness.
Add explicit validation/confirmation checkpoints for destructive and batch operations (e.g., list pins before `clear_pins()`, confirm count before `clear_analysis_zones()`, verify injection before `clear_injected()`).
Remove the redundant 'Important Rules' section or fold it into the existing 'Agent Fast Path' / 'Decision Framework' to cut repetition and reduce token cost.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — dense tables and executable code rather than concept explanations Claude already knows — but at ~770 lines it is voluminous, and the trailing 'Important Rules' section restates the 'Agent Fast Path' and 'Decision Framework' (open SSE first, fetch targeted, geocode before pinning, use signatures). It is not 3 because of this redundancy and bulk, and not 1 because it avoids explanatory fluff and assumes Claude's competence. | 2 / 3 |
Actionability | It provides abundant copy-paste-ready, executable Python with real method names and arguments (e.g., `await sb.place_pin(lat=34.05, lng=-118.24, label=..., category=..., ttl_hours=48)`), plus tables mapping each method to its return value and use case. It is not 2 because the guidance is concrete and complete rather than pseudocode or abstract direction. | 3 / 3 |
Workflow Clarity | A clear numbered 'Decision Framework' (steps 1–8) and per-feature 'rules of engagement' sequence the work well, but destructive and batch operations like `clear_pins()`, `clear_analysis_zones()` ('use sparingly'), and `clear_injected()` lack validation/confirmation checkpoints, which per the guidelines caps workflow clarity at 2. It is not 3 because feedback loops for risky ops are missing, and not 1 because the sequence is explicitly laid out. | 2 / 3 |
Progressive Disclosure | The body is well organized into clear sections, but it is a single ~770-line monolithic SKILL.md with no bundle files and no references — the enriched-field tables, full 20-entry signature table, and detailed GT/SAR method tables are reference material that should live in separate files. It is not 1 because the sectioning prevents a 'wall of text' feel, and not 3 because no content is split out into one-level-deep references. | 2 / 3 |
Total | 9 / 12 Passed |