Content
86%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.
A strongly engineered skill body: concrete commands everywhere, hard anti-fabrication rules with cited verification dates, explicit validation gates for the risky authenticated paths, and exemplary file splitting with a single well-signaled reference. Its only real costs are a motivational intro that pads the opening and a compressed diagnostics section whose full sequence lives in the reference.
Suggestions
Trim the two-sentence motivational opening ("Engagement numbers are the backbone... no excuse to invent them") and the repeated "the whole point of the skill" rationale to a single line, moving that motivation into the description where it already lives.
In "Diagnose access before downloading or ASR", add a short numbered sequence (probe → read report → check download_allowed/asr_allowed → stop or proceed) so the gate order is readable without opening the reference contract.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious operational knowledge (proxy/UA/412 quirks, CJK collation trap, login-gating evidence, entitlement gates), but the motivational intro ("Engagement numbers are the backbone of any honest 'why did this do well' analysis... no excuse to invent them") and the repeated "the whole point of the skill is that the number is cheap to fetch" rationale are over-explanation that could be trimmed — anchor 4 rather than the every-token-earns-its-place of anchor 5. | 4 / 5 |
Actionability | Every section ships copy-paste commands with argument forms ("scripts/bili-fetch.sh BV1xxxxxxxxx", "scripts/bili-danmaku.sh BV1xxxxxxxxx [P]", "scripts/bili-subs.sh BV1xxxxxxxxx chrome", "scripts/bili-access.py probe --bvid ... --page 1"), plus the exact JSON output shape and a runnable unittest command — fully executable coverage of the common cases. | 5 / 5 |
Workflow Clarity | Explicit validation checkpoints exist (self-test PASS/FAIL rows, "consume asr_allowed only after verify-report", "Stop on paid preview, denied or unknown entitlement", and a fail→fix→re-confirm→update-dates maintenance loop), but the overall flow is assembled across sections rather than stated, and the dense diagnostics section defers its full command sequence to the external contract — minor gaps consistent with anchor 4 rather than anchor 5. | 4 / 5 |
Progressive Disclosure | Clean split: executable quick start and output contract inline; the full endpoint catalog, WBI signing, and gotchas are one level deep in references/bilibili_api.md with descriptive link text and a verified anchor link; scripts are explicitly marked "execute (don't read them as reference)" — a clear overview with well-signaled references. | 5 / 5 |
Total | 18 / 20 Passed |