Content
50%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 a reasonably actionable research playbook with concrete grep/glob and MCP examples and a clear four-step methodology, but it is padded with self-evident best-practices and persona boilerplate, lacks validation checkpoints in its workflow, and keeps all detail in a single monolithic file.
Suggestions
Cut the 'Best Practices' list and the opening persona sentence; Claude already knows to be thorough and document findings, so these are pure padding competing for context.
Add explicit validation/feedback steps to the research workflow, e.g. after gathering, verify findings against a second source or confirm dependency claims by running an import check before recording them.
Move the YAML output schema and the MCP tool-call reference into separate reference files (e.g. OUTPUT_SCHEMA.md, MCP_TOOLS.md) and link to them from a concise Quick Start overview to improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably verbose — long boilerplate like the opening 'You are a research specialist focused on thorough investigation...' and self-evident Best Practices ('Be Thorough', 'Stay Organized', 'Document Everything') restate what Claude already knows, though the concrete command snippets do earn some efficiency. | 2 / 5 |
Actionability | It provides mostly executable guidance — real grep/glob command examples and concrete MCP tool-call snippets with arguments — with only minor gaps such as the 'Cross-Reference' section's malformed 'class$function' token and pseudocode-ish search strategies. | 4 / 5 |
Workflow Clarity | A numbered methodology (Information Gathering → Pattern Analysis → Dependency Analysis → Documentation Mining) gives a clear sequence, but there are no validation checkpoints or feedback loops despite coordination/memory writes that could fail silently; checkpoints are only implicit. | 3 / 5 |
Progressive Disclosure | Content is organized into clear sections with a defined YAML output schema, but everything is inlined into one ~170-line file with no external references or bundle files, so there is room to split the API/MCP details and the output schema into separate reference files. | 3 / 5 |
Total | 12 / 20 Passed |