Content
63%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 tightly written instruction-only skill with genuinely actionable interview scripts, a concrete brief template, a routing table, and a pre-delivery validation screen. Its weaknesses are redundancy around loading the Director's Read, a flow whose core classification step depends on reference files that live outside the bundle and are not present, and the absence of any bundled reference material to back the heavy external linking. It lands solidly above the midpoint but short of excellent.
Suggestions
Consolidate the three separate "load the Director's Read" instructions into one directive at the top of Process, cutting the duplicated framing sentences in the opening paragraphs.
Bundle the referenced files (directors-read.md, interview-starters.md, moderation-prescreen.md) inside this skill's own references/ directory and update links to relative paths within the bundle, so the skill is self-contained and navigable.
Trim the Intent section's non-operational prose ("Speed is the form their trust takes") and add a one-line recovery instruction for the Pre-Delivery Screen — what to do when the pre-screen flags the draft — to close the validation loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with domain-specific directives and explains nothing Claude already knows, but it repeats the same instruction three times — "load the [Director's Read]" appears in the opening, again two paragraphs later ("Load the [Director's Read] before producing the compact brief"), and once more in Process — and the Intent section carries non-operational prose ("The soul of this skill is restraint... Speed is the form their trust takes"). Anchor 4's "minor instances that could be trimmed" understates this redundancy, while anchor 1-2 verbosity (padding, known-concept explanation) is absent. | 3 / 5 |
Actionability | The skill provides copy-paste-ready interview questions with defaults attached ("What happens in the video... (not sure? I'll pick one simple action with a visible ending)"), a fill-in brief template ("Mode: [T2V/I2V/V2V/R2V]. Subject: [anchor]..."), and an explicit routing table with named skills and conditions. It falls short of anchor 5 because the central workflow step — "classify the brief, and complete its canonical narrative or non-narrative record" — depends entirely on the Director's Read file, and none of the referenced files exist in this bundle, so the core action cannot be executed from the skill alone. | 4 / 5 |
Workflow Clarity | The flow is legible across sections: load the Director's Read and classify, run the ≤3-question interview, fill the Compact Brief Pattern, run the Pre-Delivery Screen, then route. An explicit validation checkpoint exists ("Run the [moderation pre-screen] on the draft before it is shown"), plus the accepted-footage gate for sequences. It misses anchor 5 because the pre-screen has no feedback loop (what to do when the screen flags the draft is only sketched as "keep the scene, drop the cue") and steps are not explicitly sequenced against each other. | 4 / 5 |
Progressive Disclosure | The body is well-sectioned with clearly signaled links, but every detailed reference — directors-read.md, interview-starters.md, directing-engine.md, moderation-prescreen.md — points to "../../references/", outside the skill's own bundle, and the referenced sibling skills (../seedance-sequence/SKILL.md, etc.) are likewise absent from this workspace. The bundle contains no references/, scripts/, or assets/ directories at all, so navigation from this skill alone is broken and the skill is not self-contained. That is a larger structural gap than anchor 4's "minor organization gaps", though better organized than anchor 3's typical buried-reference case. | 3 / 5 |
Total | 14 / 20 Passed |