Content
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body delivers concrete, well-sequenced browser workflows with explicit validation and error-recovery loops, but is held back by repeated restatement of the same session-boundary constraints, a couple of placeholder selectors in the core workflows, and an absence of any progressive-disclosure structure for its substantial inline reference material.
Suggestions
State the '不开独立 session、不 import cookie' and fail-first/session-busy constraint once in a dedicated 约束 section and reference it instead of repeating it verbatim across the intro, 互动流程, mode note, pitfalls, and error table.
Replace the '<send-btn-selector>' placeholder in the comment workflow and the 'camoufox-cli ... open' ellipsis in the follow workflow with concrete selectors (or a snapshot-based discovery step) so the core actions are copy-paste ready.
Move the Pitfalls and 错误处理 reference material into a references/ file (e.g. PITFALLS.md) linked from the workflow section, so the main body stays a lean overview with one-level-deep navigation.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational, but the boundary constraint '不开独立 session、不 import cookie' and the fail-first/session-busy explanation are restated verbatim across the intro, 关键边界, 互动流程, mode note, pitfalls, and error table, which is repeatable verbosity that could be tightened. It is not a 1 because it does not explain concepts Claude already knows, and most operational detail earns its place; not a 3 because the redundancy is clear. | 2 / 3 |
Actionability | Most steps give concrete executable camoufox-cli commands with real selectors (e.g. type '.content-input', eval classList checks), but the core comment step uses a placeholder 'click "<send-btn-selector>"' and the follow section uses 'camoufox-cli ... open' with an ellipsis, leaving key details incomplete. It is not a 1 because the majority of commands are concrete and runnable; not a 3 because the send-button selector and the abbreviated follow command are not copy-paste ready. | 2 / 3 |
Workflow Clarity | Each interaction is a clearly numbered sequence with explicit validation checkpoints (snapshot confirms the comment/reply appeared, eval verifies like-state) and feedback loops (retry once then report, wait 60s and retry on the same session, re-login-manager flow on cookie expiry), so batch comment operations are not left without verification. It is not a 2 because checkpoints are explicit rather than implicit or missing. | 3 / 3 |
Progressive Disclosure | The content is well-organized into sections but lives entirely in one ~220-line SKILL.md with no bundle files and no one-level-deep references, and substantial reference material (eight pitfalls plus the error-handling table) is inline rather than split out. It is not a 1 because the sections are clearly headed rather than a monolithic wall and there are no nested references; not a 3 because nothing is split across files and the skill exceeds the under-50-lines simple-skill allowance. | 2 / 3 |
Total | 9 / 12 Passed |