Content
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A thorough, safety-focused connector guide with clear sequencing and concrete output artifacts, but it is verbose and repetitive, lacks validate→fix→retry feedback loops for batch/destructive work, and is a monolithic single file with no progressive disclosure into bundle references.
Suggestions
Consolidate the repeated safety rules (credit exposure, mutation gates, sequence launch gates) into a single Safety Rules section and reference it elsewhere to remove redundancy and tighten conciseness.
Add an explicit validate→fix→retry feedback loop for batch and mutation operations (e.g. after a bulk create-update, verify the returned records, and if duplicates or errors appear, fix and re-run) to lift workflow clarity.
Split detailed material (per-domain table schemas, the output-format template, identifier rules) into reference files under references/ and link to them one level deep, and fix the malformed "##" anchor in the Cross-Skill Best Practices link.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The guidance is domain-specific (credit, version, mutation rules) rather than explaining concepts Claude already knows, but safety admonitions are restated across Credit-Aware, Mutations, Sequences, and Safety Rules sections, so it could be tightened; this fits the mostly-efficient-but-could-be-tightened anchor rather than the lean anchor at 3. | 2 / 3 |
Actionability | Concrete artifacts exist (output-format template, exact table column specs, field defaults like "Keep `reveal_personal_emails` false"), but much guidance is abstract and conditional ("use the live Apollo tool schema", "when they improve precision", "when exposed"), leaving it mixed rather than uniformly copy-paste ready. | 2 / 3 |
Workflow Clarity | Workflows are clearly sequenced (Apollo Operation Ranking 1–6, People-to-Company-Details 1–7) with explicit preview/approval/verification checkpoints, but there is no explicit validate→fix→retry feedback loop for the batch and destructive operations, which caps the score at 2 per the rubric's feedback-loop guidance. | 2 / 3 |
Progressive Disclosure | Section headers are well-organized and external references to ../index/SKILL.md are one level deep and signaled, but no bundle files exist and the skill is a monolithic ~210-line single file with inline table schemas, output format, and per-domain rules that could be split; additionally one anchor is malformed ("SKILL.md## Cross-Skill Best Practices" uses "##" instead of "#"). | 2 / 3 |
Total | 8 / 12 Passed |