Content
85%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-engineered complex skill: a validated five-phase workflow, concrete scoring rubrics and output specs, and clean one-level progressive disclosure across ten real reference files. The main weakness is token efficiency — three sections (inputs vs questions, two tool tables, repeated output rules) duplicate each other and could be consolidated without losing anything.
Suggestions
Merge "Inputs to Collect" and "Task-Specific Questions" into one section — both enumerate the same seven items (branch, ICP, count, geography, tools, format, signal preference).
Consolidate "Tool Selection Quick Picks" and "Tool Integrations" into a single table (tool, use, MCP, guide link), or move the integrations table into references/data-sources.md — ~10 tools currently appear in both tables.
State the "Top outreach targets" (top 3–5 hot leads with rationale) requirement once — it is currently specified in both Phase 5 and "Always include after the table".
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The prose is lean and doesn't explain concepts Claude already knows, but there is real duplication: "Inputs to Collect" and "Task-Specific Questions" ask nearly the same seven items, "Tool Selection Quick Picks" and "Tool Integrations" overlap on ~10 tools, and the "Top outreach targets" rule appears in both Phase 5 and Output Formats. This is 'could be tightened' territory rather than minor trims, so it sits at the anchor-3 level rather than 4. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is fully executable: exact phase sequence, per-branch source lists, Hot/Warm/Cold/Skip scoring definitions, confidence criteria ("High: confirmed by at least two independent sources"), exact chat-table and CSV column specs, concrete defaults (25/15 target counts, 2–3× oversourcing, re-verify within 7 days), and an invalid-email bucket rule. An agent could execute this without guessing. | 5 / 5 |
Workflow Clarity | The five phases are clearly sequenced with an explicit gate ("Don't move to discovery without this"), validation checkpoints (email verification before the list and within 7 days of sending, confidence levels), an 8-item pre-finalization Quality Checks checklist, and error-recovery paths (invalid bucket, "Open questions", explaining a smaller final count). This matches the anchor with explicit validation, feedback handling, and a checklist for a batch operation. | 5 / 5 |
Progressive Disclosure | The body is a clear overview with all branch deep-dives and specialized content (scoring signals, compliance detail, enrichment playbook, data sources, account research) split into 10 one-level-deep reference files, each existing on disk and clearly signaled at its point of use with a description of what it contains. No nested reference chains within the skill. | 5 / 5 |
Total | 18 / 20 Passed |