Content
87%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill body is actionable and concise, with concrete MCP tool calls and a reusable report template. Its main weakness is workflow clarity: validation and cross-referencing controls are stated separately rather than embedded as checkpoints in the research sequence.
Suggestions
Fold the Quality Rules (source-every-claim, cross-reference, recency, acknowledge gaps) into the workflow as explicit verification checkpoints — e.g. a Step 5.5 'Verify' pass before writing the report.
Add a feedback loop for the synthesis step: if a sub-question has no sourced findings, retry with alternate keyword variations before flagging 'insufficient data' in the report.
Trim or relocate the Examples block to a reference file or condense it, since it largely restates triggers already covered in When to Activate.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean: it jumps straight into workflow steps with exact tool signatures and a report template, never explaining what research or MCPs are; the drift warning earns its tokens by flagging time-sensitivity. | 3 / 3 |
Actionability | Concrete executable tool calls with parameters (e.g. firecrawl_search(query, limit: 8), crawling_exa(url, tokensNum: 5000)), a copy-paste report template, and an explicit parallel-subagent plan make guidance copy-paste ready. | 3 / 3 |
Workflow Clarity | The 6 steps are clearly sequenced, but quality/cross-check controls live in a separate "Quality Rules" section rather than as inline checkpoints in the workflow, so verification is present but implicit rather than wired into the sequence. | 2 / 3 |
Progressive Disclosure | A single-file skill with no bundle files and clean section organization (When to Activate, MCP Requirements, Workflow, Quality Rules, Examples); the drift warning correctly surfaces time-sensitive caveats inline. | 3 / 3 |
Total | 11 / 12 Passed |