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 rich in executable, copy-paste-ready commands and contains genuinely good collaboration patterns (Think-Act-Observe, iterative decomposition, dry-run preview). However, it is severely overlong for a SKILL.md, repeats its own safety guidance multiple times, and its closing Examples showcase exactly the risky one-shot YOLO workflows the skill itself forbids.
Suggestions
Cut the body to a lean overview (~100 lines): keep install/auth/basic usage, one YOLO safety table, and the Think-Act-Observe pattern; move configuration, troubleshooting, and the examples into references/ files linked with 'See [X.md](X.md)' navigation.
Rewrite or delete the four 'Examples' one-shot YOLO scripts — they are batch/destructive workflows with no validation steps and directly contradict the 'NEVER Use YOLO For' guidance; replace them with the iterative Claude-orchestrated pattern already demonstrated in Sequential Task Decomposition.
Deduplicate YOLO guidance into a single section and consolidate the 'When to Use', 'Best Practices', and 'Tool Execution' sections, which currently repeat the same safety advice three times.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 660-line body is noticeably padded: the 'When to Use' section repeats the frontmatter description, YOLO safety guidance is restated across 'YOLO Mode', 'Best Practices > Security', and 'Tool Execution', and the four 'Examples' are near-duplicate giant one-shot prompt walls. It avoids anchor 1 only because it does not explain concepts Claude already knows — the padding is repetition, not education — and it embeds time-sensitive specifics (gemini-2.5-flash/pro, '60 req/min, 1000 req/day') outside any deprecation section. | 2 / 5 |
Actionability | Nearly all guidance is copy-paste-ready bash — install ("npm install -g @google/gemini-cli"), auth ("export GEMINI_API_KEY=..."), usage flags, a complete settings.json example, and troubleshooting commands. Falls short of anchor 5 due to minor gaps: `gemini mcp add` is shown with no arguments, `claude-cli` is assumed as an available binary, and `gemini changelog`/`gemini usage` are unverified commands. | 4 / 5 |
Workflow Clarity | The Think-Act-Observe loop and Sequential Task Decomposition patterns show clear sequencing with explicit review checkpoints and an error-handling feedback loop, but validation is inconsistent: the 'Examples' section presents destructive/batch workflows ('Daily Development Automation' ending in 'Commit and push changes', full framework migration) as single YOLO commands with zero verification steps, contradicting the body's own 'NEVER Use YOLO For' list. Per the rubric's cap, batch operations without validation cap this at 3 even though earlier sections sequence well. | 3 / 5 |
Progressive Disclosure | A single 660-line file with no bundle files at all: reference material that clearly belongs in separate files (settings.json configuration, environment variables, troubleshooting, the Examples section) is fully inlined. Clear ## / ### headers provide real structure (above anchor 2's 'no section headers' example), but the monolithic sizing and a 'Related Skills' list pointing to unverifiable skills keep it below anchor 4. | 3 / 5 |
Total | 12 / 20 Passed |