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.
A well-structured, actionable skill body that sequences a paid batch workflow with appropriate validation and approval checkpoints. It earns its token budget with concrete commands and JSON schemas rather than conceptual filler.
Suggestions
Add an explicit 'If the --test run fails or returns irrelevant authors: refine keywords/exclusions, then re-run --test' feedback loop so error recovery is spelled out rather than implied.
Move the config JSON schema and output CSV column table into a references/ file (e.g. CONFIG.md) to tighten SKILL.md and deepen progressive disclosure.
De-duplicate the invocation command — the Example Usage block repeats the full command; reference the flags table instead.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and phase-structured with no padding about what LinkedIn/KOLs are; the only minor trim opportunity is the repeated full command in the Example Usage section. | 4 / 5 |
Actionability | Provides a copy-paste bash command with every flag documented, a concrete config JSON schema, and the web-kols JSON format; minor gap is that the agent-generated web-kols path is only loosely specified. | 4 / 5 |
Workflow Clarity | Phases 0–4 are clearly sequenced with validation checkpoints ('Always run with --test first', 'Present keywords to user for approval before running') and a Phase 3 refine feedback loop; just short of a 5 because error-recovery on a failed test run is implicit rather than spelled out. | 4 / 5 |
Progressive Disclosure | Well-organized into labeled phases with bulk logic delegated to the verified scripts/kol_discovery.py (one level deep); minor gap is that the config schema and CSV column reference are inlined rather than split into a reference file. | 4 / 5 |
Total | 16 / 20 Passed |