Content
77%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a well-organized, actionable routing skill with concrete identifiers, commands, and a clear sequenced workflow, but it suffers from redundant restated sections and references refs/*.md files that are not present in the bundle.
Suggestions
Consolidate the overlapping decision sections ('执行前四问', '什么时候用', '不要用在这里', '默认路由顺序', '路由矩阵', '和其他 skill 的区别') into a single routing flow to remove repetition and save tokens.
Add the missing refs/*.md bundle files (playwright-mcp.md, playwriter.md, agent-browser.md, pinchtab.md) that the body repeatedly routes to, so navigation actually resolves.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is table-driven and free of concept-explanation fluff, but the same routing decision is restated across overlapping sections ('执行前四问', '什么时候用', '不要用在这里', '默认路由顺序', '路由矩阵', '和其他 skill 的区别'), which could be tightened. | 2 / 3 |
Actionability | Guidance is concrete and copy-paste ready: specific MCP IDs, tool prefixes ('mcp__claude-in-chrome__*'), commands ('npx @playwright/mcp@latest', 'npm i -g agent-browser', 'pinchtab mcp'), ref paths, and a specific failure fix ('pinchtab_eval + window.location.href' for the Clash TUN 403). | 3 / 3 |
Workflow Clarity | A clear 7-step '默认路由顺序' with explicit '是 → X' branches, a four-question pre-flight checklist, a '交付要求' reporting checklist, a liveness-check note ('使用前先做本机验活'), and a Common Mistakes error-recovery table provide a well-sequenced, checkpointed workflow. | 3 / 3 |
Progressive Disclosure | The overview is well-signaled toward one-level-deep refs ('读取哪些 refs' table points to refs/playwright-mcp.md etc.), but those refs/*.md files do not exist in the bundle, so the referenced navigation paths are broken rather than functional. | 2 / 3 |
Total | 10 / 12 Passed |