Content
27%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill is significantly over-engineered and verbose for what it accomplishes. It explains research methodology and concepts that Claude already understands (how to search for companies, what products and services are, what B2B means). The report template is the most valuable part, but it's buried in ~250 lines of largely redundant guidance. The content would benefit enormously from being reduced to the template plus a few key constraints, with any detailed variations split into separate files.
Suggestions
Cut the content by 60-70%: remove the 'When to Use' section (covered by YAML), the research methodology phases (Claude knows how to research), and obvious advice like 'don't include outdated information.' Focus on the report template and any non-obvious constraints.
Extract the report template, search patterns, and report variations into separate referenced files (e.g., REPORT_TEMPLATE.md, SEARCH_PATTERNS.md, VARIATIONS.md) to improve progressive disclosure.
Make subagent delegation actionable: provide actual invocation syntax or tool-use examples for 'web-researcher' and 'people-finder' instead of vague descriptions.
Add a concrete feedback loop for handling missing information: e.g., 'If a section cannot be populated after 3 search attempts, mark it as [Not publicly available] and note the last verified date.'
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Extremely verbose at ~250+ lines. Explains obvious concepts Claude already knows (what B2B/B2C means, how to search the web, what a company profile contains). Time estimates like '5 minutes' and '10-15 minutes' are meaningless for Claude. The 'When to Use This Skill' section restates what the YAML description already covers. Much of this is generic research methodology that Claude inherently understands. | 1 / 3 |
Actionability | The report template is concrete and copy-paste ready, which is valuable. However, the 'bash' search queries are not actual executable commands—they're just search strings. The research methodology is more of a checklist of things to look for rather than specific executable steps. References to subagents like 'web-researcher' and 'people-finder' lack specifics on how to invoke them. | 2 / 3 |
Workflow Clarity | The three-phase structure (Discovery → Detailed Research → Synthesis) provides a clear sequence, and the quality checklist at the end serves as a validation step. However, there are no feedback loops or error recovery steps—e.g., what to do when information is unavailable for key sections, or how to handle conflicting sources. The phases have arbitrary time estimates rather than meaningful checkpoints. | 2 / 3 |
Progressive Disclosure | This is a monolithic wall of text with no bundle files and no references to external documents. The report template, search patterns, report variations, and detailed methodology are all inlined when they could be split into separate reference files. At this length, the skill desperately needs a concise overview with pointers to detailed materials. | 1 / 3 |
Total | 6 / 12 Passed |