Content
67%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 with executable commands, a complete schema, error-recovery guidance, and explicit done criteria. Its main weaknesses are padding in the server-side taxonomy section, an inconsistent and unresolvable script path, and no per-item validation for the batch pattern.
Suggestions
Trim the eight-bullet virality-framework breakdown (or move it to a reference file) — the agent cannot tune those knobs and the ranking rationale already comes back in the output's virality_reason field.
Use one consistent, existing script path throughout (Step 3 uses 'library/edit/ai-clipping/scripts/run-ai-clipping.sh' while the patterns call a bare 'run-ai-clipping.sh') and ship the script in the bundle.
Add a validation checkpoint for the batch pattern (e.g., check each run's exit status or result count before continuing with xargs).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient tables and commands, but the eight-bullet virality-framework taxonomy under 'What Happens Server-Side' explains server internals the agent cannot influence (redundant with mistake #4, "those knobs aren't exposed"), and the 'no ffmpeg / no local Whisper / no GPU' point is repeated across the intro and Step 2. This matches the 'mostly efficient but includes some unnecessary explanation' anchor, below level 4's 'minor instances'. | 3 / 5 |
Actionability | The skill provides copy-paste-ready bash invocations, a complete input flag table, a concrete JSON output schema, and per-case patterns (defaults, batch, async, local file, coords-only). It is not a 5 because the script path is inconsistent — "library/edit/ai-clipping/scripts/run-ai-clipping.sh" in Step 3 versus bare "run-ai-clipping.sh" in the patterns — and the script file does not exist in the bundle. | 4 / 5 |
Workflow Clarity | A clear sequence exists (collect inputs → verify prerequisites → run → done criteria) with failure-mode recovery loops (re-fetch by request_id, bump --poll-timeout, upload then retry). It falls short of 5 because the batch pattern (xargs -a urls.txt) lacks per-item failure validation and verification is implicit ('the file exists and parses') rather than an explicit validate step. | 4 / 5 |
Progressive Disclosure | The single-file body is well-organized with clear, well-signaled sections (When to Use, Execution Protocol, Output Schema, Failure Modes, Done Criteria) and no nested references, which suits a single-purpose API-wrapper skill. Minor gaps: the referenced script is absent from the bundle, and content like the aspect-ratio picker and server-side taxonomy could live in a reference file. | 4 / 5 |
Total | 15 / 20 Passed |