Content
77%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 highly actionable and workflow-safe: routing is unambiguous, error handling is concrete and two-layered, and risky operations (notes writes, PS 5.1 encoding) are guarded by explicit validation. Its weaknesses are duplicated error/rule explanations that inflate token cost, and a bundle in which every referenced file is missing while detailed platform rules are inlined rather than separated.
Suggestions
Deduplicate the error-code documentation: the -100/-200 semantics appear in both the API template section and the SKILL Update section — keep one canonical explanation and reference it elsewhere.
Move the Detailed UTF-8 Encoding Rules and PowerShell 5.1 sections into a reference file (e.g. references/encoding.md) and keep only the one-line mandatory rule in SKILL.md, which would also improve progressive disclosure.
Ensure the referenced bundle files actually ship with the skill (notes/SKILL.md, knowledge-base/SKILL.md, ima_api.cjs, meta.json) — none were present in the evaluated bundle, so the routing table and API template currently point at nonexistent paths.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly operational (tables, code, checklists), but the -100/-200 error-code semantics are explained twice (in the API template blockquote and again in the SKILL Update section), the UTF-8 rule is stated in the mandatory rules and then re-stated at length inline, and the PowerShell section repeats its warning multiple times ("这是 PowerShell 5.1 的已知设计缺陷，不是 bug 可以被修复"). This matches anchor 3 (mostly efficient but could be tightened), not 4 since the repetition is more than minor. | 3 / 5 |
Actionability | Fully executable, copy-paste-ready guidance throughout: a complete bash API-call template with two-layer error capture and jq parsing, credential setup commands for config-file and env-var modes, and working transcoding snippets for Python, Node.js, iconv, and PowerShell, plus a PS version-detection script. Matches anchor 5. | 5 / 5 |
Workflow Clarity | Multi-step flows are explicitly sequenced with validation checkpoints: credential check before any API call, a module decision table for routing, a mandated read-order for cross-module tasks ("先读 knowledge-base/SKILL.md → 再读 notes/SKILL.md"), a UTF-8 validation checklist before every notes write, and a -200 feedback loop (read update context → guide user → retry request). Matches anchor 5 (explicit validation steps, feedback loops, checklists). | 5 / 5 |
Progressive Disclosure | The routing table clearly signals which submodule SKILL.md to read per intent (one level deep), but the bundle contains none of the referenced files (no notes/ or knowledge-base/ SKILL.md, no ima_api.cjs, no meta.json), and ~100 lines of platform-specific detailed rules (UTF-8 transcoding, PowerShell 5.1) are inlined in SKILL.md instead of being split into reference files. This sits between anchor 3 (content that should be separate is inline) and anchor 4, landing at 3. | 3 / 5 |
Total | 16 / 20 Passed |