Content
65%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 exceptionally actionable — nearly every capability ships with executable code and concrete arguments — and the fast-path section is a strong navigation device. It is weakened by monolithic inlining of reference tables that belong in separate files, redundant restatement of the same rules, and no validation checkpoints around destructive/batch operations. Splitting reference material out and adding confirm-before-clear steps would lift it substantially.
Suggestions
Move the stable reference tables (Enriched Data Fields by Layer, GT analytics commands, pin categories/colors, message signature table, osint_lookup tool table) into references/ files (e.g. references/api.md, references/signatures.md) and keep SKILL.md as a concise overview with one-level-deep links.
Add explicit validation/confirmation steps before destructive and batch operations — e.g. list matching pins with get_pins() and confirm with the operator before clear_pins(), and validate batch payloads (count, categories) before place_pins_batch()/send_batch().
Deduplicate the latency tiers, anti-patterns, and 'fetch targeted, not everything' guidance that appear in both 'Agent Fast Path' and 'Important Rules', stating each rule once and cross-referencing.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~780-line body is mostly dense API surface rather than general-knowledge padding, but it repeats itself: latency tiers and anti-patterns appear in both 'Agent Fast Path' and 'Important Rules', 'When to use' notes duplicate the Decision Framework, and the analysis-zone prose ('Think of them as sticky notes...') is padded. This matches anchor 3 — mostly efficient but could be tightened — rather than 4, where only minor trimming would be needed. | 3 / 5 |
Actionability | Executable, copy-paste-ready code throughout: 'from sb_query import ShadowBrokerClient; sb = ShadowBrokerClient()', 'await sb.place_pin(lat=34.05, ..., confidence=0.8, ttl_hours=48)', 'python openclaw-skills/shadowbroker/verify_hmac.py', plus complete argument tables for osint_lookup. Specific examples cover the common cases, matching the score-5 anchor. | 5 / 5 |
Workflow Clarity | The Decision Framework gives a clear numbered sequence with some checkpoints ('Open SSE stream first', 'geocode before pinning', 'Call sar_status() first', and the verify_hmac.py verification step), but destructive and batch operations (clear_pins(), clear_injected(), clear_analysis_zones(), place_pins_batch, send_batch) carry only 'use sparingly' advice with no validation or confirmation step. Per the rubric guideline, missing validation in destructive/batch workflows caps this dimension at 3. | 3 / 5 |
Progressive Disclosure | No references/, scripts/, or assets/ bundle exists, so the entire tool surface is inlined in one 783-line file: the 'Enriched Data Fields by Layer' table, the GT analytics command table, the pin-category color table, and the 20-row signature table are reference material that clearly belongs in separate files. Structure (headings, a 'Fast Path (read first)' section, tables) is better than the 'minimal structure' of anchor 2, but inline reference content and the absence of any split files keep it below anchor 4. | 3 / 5 |
Total | 14 / 20 Passed |