Content
70%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 highly actionable skill with an excellent fallback-aware workflow, undermined by redundancy and monolithic structure. The model-override documentation and MCP setup instructions are duplicated, and one-time setup material plus the prompt template belong in reference files to slim the main body.
Suggestions
Move the Environment & Setup section (prerequisites, MCP configuration variants, authentication) into a references/SETUP.md file and keep a one-line pointer plus the verification command in SKILL.md, cutting ~70 lines from the always-loaded body.
Consolidate the model override documentation: the DEFAULT_MODEL bullet and the override blockquote list repeat the same options — keep one canonical table or list with the syntax, effect, and caveat per option.
Place the full search prompt template in a reference file (or a single canonical block) and make the CLI fallback reference it explicitly instead of "...same prompt as above...", and embed the timeout directly in the CLI command (e.g., timeout 120 gemini -p '...').
Trim the "When to use which" subsection and the three MCP setup variants (JSON, npx JSON, one-liner) down to the one-liner plus one config example, since the "MCP first, CLI second" rationale already appears in Key Rules.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Operational specifics Claude would not know (MCP tool names, model routing quirks, stderr/timeout handling) earn their tokens, but there is noticeable redundancy: model overrides are documented twice (Constants bullet plus the blockquote list), three MCP setup variants are given, and "When to use which" repeats the "MCP first, CLI second" rule already in Key Rules. | 3 / 5 |
Actionability | The full mcp__gemini-cli__ask-gemini call with a copy-paste prompt template, exact MCP JSON configs, npm install commands, and a CLI fallback command are highly executable; minor gaps include the CLI fallback eliding the prompt ("...same prompt as above...") and the 120s timeout being stated as prose rather than embedded in the command. | 4 / 5 |
Workflow Clarity | Five clearly sequenced steps with explicit error-recovery loops (MCP failure falls back to CLI, then to alternative skills), a Verify Setup check, and degraded-output handling ("If Gemini returns fewer papers than requested, note this but do not re-query") match the explicit-validation-and-feedback-loops anchor; no destructive or batch cap applies. | 5 / 5 |
Progressive Disclosure | Sections are well-headered and navigable, but roughly 70 lines of one-time setup/install/auth content and the long search prompt template are inlined in a ~225-line SKILL.md with no reference files, fitting the anchor where content that should be separate is inline. | 3 / 5 |
Total | 15 / 20 Passed |