Content
75%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 a disciplined, policy-dense overview that routes correctly to real workflow references with strong safety checkpoints around billing, consent, and failure handling. Weaknesses are mild: some duplicated rules, and two bundled reference files left unlinked from the main body.
Suggestions
State the resolved-URL download-failure rule once (in the capabilities block) and drop its duplicate in 'Files and results'.
Add one-line pointers to references/agent-compatibility.md and references/api-contract.md with their read conditions, so all bundle files are navigable from SKILL.md.
Break the longest compound sentences (e.g. the route-selection and cancellation rules) into shorter imperative sentences to reduce re-reading cost.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body contains no explanations of concepts Claude already knows and nearly every sentence carries an operational rule (URLs, env var names, param names), so it is well above the verbose anchors. It misses anchor 5 because of repetition and dense multi-clause sentences: the resolved-URL download-failure rule appears in both the capabilities block ("If the agent cannot download the resolved video URL... give the user the resolved data.video URL") and again in Files and results ("If resolved video downloading fails, return the resolved data.video URL"), and 'a share link is not a substitute for a local video' is likewise stated twice. | 4 / 5 |
Actionability | Concrete, executable elements are present: the MCP endpoint "https://www.550wai.cn/mcp/global", a full stdin JSON example for the CLI fallback ("node {baseDir}/dist/550w-skill.cjs" with {"action":"queryCredits",...}), exact env vars, and the confirmProcessing=true parameter semantics. It is below anchor 5 because the primary upload/submit/poll request shapes are not shown inline — they are delegated to the workflow references — so the body alone leaves minor gaps for the most common operations. | 4 / 5 |
Workflow Clarity | Route selection is a clear numbered sequence with explicit checkpoints for the risky (billed/destructive) operations: consent plus confirmProcessing=true before paid submission, "check the original task before doing anything else; never resubmit" on timeout, and 'Task deletion requires separate user approval', so the destructive-operation cap at 3 does not apply. It stops short of anchor 5 because the end-to-end media processing sequence (inspect → prepare → upload → submit → poll) is not laid out in the body — it lives one level deep in the references — and checkpoints are scattered across prose sections rather than one coherent checklist. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview that cleanly routes to one-level-deep, existing references ([OAuth workflow](references/oauth-access.md) and [API Key workflow](references/openapi-access.md), both present in the bundle) and mentions scripts/550w-upload.cjs; the detailed numbered workflows are appropriately split out rather than inlined. It misses anchor 5 because two bundle references, references/agent-compatibility.md and references/api-contract.md, are never linked or named from SKILL.md — they are discoverable only by listing the directory — leaving a minor navigation gap. | 4 / 5 |
Total | 16 / 20 Passed |