Content
87%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-engineered, highly actionable hunting skill body: concrete 5-phase workflow with validation gates and compliance redlines, real verified bundle files behind clear one-level links, and specific tools/payloads rather than abstractions. Minor conciseness and progressive-disclosure gaps come from the inlined inventory table and reference-tree depth.
Suggestions
Trim or relocate the '数据资产规模' inventory table into a reference file so the entry body stays lean; keep only the counts Claude needs for routing.
Note explicitly in the tree diagram that h1-reports/ and payloader/raw JSON are machine-indexed data (not to be read wholesale), to reinforce progressive-disclosure guidance and prevent context blowup.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and mostly token-efficient — tables, short bullets, and cross-links assume Claude's competence rather than explaining basics — but it carries some inventory padding (the '数据资产规模' table and large reference tree) that could be trimmed, keeping it just below the lean/efficient top anchor. | 4 / 5 |
Actionability | Provides concrete executable guidance throughout: named tools (amass/subfinder/httpx/ffuf/ffuf), exact dorks (`site:target.com inurl:/admin`), specific file pointers per attack type, MCP activation calls (`mcp__jshook__search_tools`), and a copy-ready report skeleton with CVSS 4.0 vector requirement. | 5 / 5 |
Workflow Clarity | A clearly sequenced 5-phase workflow (intake→recon→enum→hunt→report) with explicit validation checkpoints: an ACTION REQUIRED gate (confirm scope.md exists and auth.status=granted), priority routing tied to hit-rate data, compliance redlines as 'don't do' gates, and a completion self-check checklist with feedback loops for a destructive/batch context. | 5 / 5 |
Progressive Disclosure | SKILL.md is an overview with one-level-deep, clearly signaled markdown links to verified real files (methodology/, playbooks/, industry/, dictionaries/, templates/, tools/) and a tree diagram; it is well-organized, though the inlined asset-scale table and the slightly mixed reference/index depth keep it just below the cleanest top anchor. | 4 / 5 |
Total | 18 / 20 Passed |