Content
57%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 skill body is a well-structured, token-efficient actor-selection catalog with genuinely executable mcpc commands and a clear five-step workflow. Its main weaknesses are the missing bundle script that Step 4 depends on, the absence of any output-verification checkpoint for paid batch runs, and a large inline catalog that should live in a reference file. These cap actionability, workflow clarity, and progressive disclosure at the midpoint.
Suggestions
Ship the referenced script (place run_actor.js in the bundle's scripts/ directory and fix the path) or replace the Step 4 commands with something self-contained, so the core execution step actually runs.
Add a verification checkpoint to Step 5 (e.g., check the output file exists, is non-empty, and row count matches expectations before summarizing; retry with reduced maxItems on failure) to satisfy the batch-operation feedback-loop requirement.
Move the per-platform actor tables to a reference file (e.g., references/actors.md) and keep only the use-case selection matrix inline, which would fix both the progressive disclosure and token-efficiency concerns.
Make Step 3's 'Number of results: Based on character of use case' concrete — suggest default caps per use case and remind to set a conservative maxItems given the pricing warning.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by dense, purposeful tables and copy-paste commands with almost no explanation of concepts Claude already knows. Minor waste keeps it from anchor 5: the frontmatter description is repeated verbatim under the H1, actor IDs are duplicated across the platform, use-case, and multi-actor tables, and the trademark disclaimer adds little operational value. | 4 / 5 |
Actionability | The mcpc schema-fetch and actor-search commands are concrete and executable, but the central Step 4 run commands invoke '${CLAUDE_PLUGIN_ROOT}/reference/scripts/run_actor.js', which does not exist in the bundle (no scripts/ directory), so the primary command fails as written. 'Number of results: Based on character of use case' is also too vague to act on — concrete guidance with a missing key piece, matching anchor 3 rather than 4. | 3 / 5 |
Workflow Clarity | The five steps are clearly sequenced with a copyable progress checklist, a pricing-approval checkpoint, and an error-handling table. However, paid batch scrape runs have no output validation or feedback loop (no step verifies the result file is non-empty/well-formed before summarizing), and the rubric caps batch-operation workflows without validation at 3. | 3 / 5 |
Progressive Disclosure | Section structure is clear (workflow, per-platform tables, use-case index, error handling), but roughly 180 lines of actor catalog that belong in a one-level-deep reference file are inlined into SKILL.md, and the only referenced path (reference/scripts/run_actor.js) is broken — the script is absent from the bundle. This matches anchor 3: structure present, but content that should be separate is inline. | 3 / 5 |
Total | 13 / 20 Passed |