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 strong, well-architected skill body: clear branch selection, a five-phase workflow with real validation gates, concrete output specs, compliance guardrails, and textbook progressive disclosure into ten existing one-level-deep references. The main weakness is token efficiency — the two overlapping tool tables and the duplicated lead-sheet column specifications could be consolidated into one section without losing any guidance.
Suggestions
Merge "Tool Selection Quick Picks" and "Tool Integrations" into a single table (tool, best for, MCP availability, guide link) — the two currently repeat ~10 of the same tools with near-identical 'best for' descriptions, costing roughly 30 lines.
State the lead-sheet columns once: Phase 5's "every lead sheet includes" list and the Output Formats section's table/CSV headers cover the same ground; keep the full specs in Output Formats and have Phase 5 point there.
Trim narrative asides like "If the user's list quality bar is high, smaller is better. 25 verified leads beats 250 mostly-junk ones." to a single clause, since the ratio guidance in Phase 4 already implies the quality-over-quantity stance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient prose — no tutorials on concepts Claude already knows, and tables/checklists are dense — but there is real redundancy that could be tightened: "Tool Selection Quick Picks" (13 rows) and "Tool Integrations" (16 rows) overlap heavily on the same tools (Apollo, Clay, FullEnrich/LeadMagic, TheirStack, ZoomInfo, Hunter, Truelist, Firecrawl, Browserbase, GitHub), and Phase 5's column list ("score, business/company name, contact... confidence, last verified date") is restated nearly in full in the Output Formats section. This fits the 3 anchor (mostly efficient but could be tightened) better than 4, where trimming needs would be minor rather than structural duplication across sections. | 3 / 5 |
Actionability | Fully executable instruction-style guidance: exact CSV column headers ("score,company,domain,industry,size_band,country,signal,..."), default counts ("default 25 for SaaS / B2B, 15 for Local SMB"), a concrete switch condition ("Switch to CSV when the list is >25 rows"), a defined scoring rubric (Hot/Warm/Cold/Skip with definitions), confidence-level criteria, and per-branch tool mappings. This is an instruction-only skill and the guidance is copy-paste ready, matching the 5 anchor; not 4, since specific examples cover the common cases across all four branches with no gaps. | 5 / 5 |
Workflow Clarity | Five clearly sequenced phases (Define ICP → discovery → qualify → score → output) with explicit validation gates: "Don't move to discovery without this", "always verify deliverability before adding to the final list, and again within 7 days of sending if the list sat", and a pre-finalization "Quality Checks" checklist with re-check semantics (e.g., invalid emails "move to a separate 'invalid' bucket and flag"). The 5 anchor (explicit validation steps, feedback loops, checklists for complex processes) fits; the level below (4) would have only minor validation gaps, and this batch-list workflow has its gates fully specified. | 5 / 5 |
Progressive Disclosure | The body is a true overview that delegates depth to ten reference files, all of which exist on disk (saas-prospecting.md, b2b-prospecting.md, local-prospecting.md, demand-signals.md, sourcing-and-enrichment.md, signal-plays.md, account-research.md, compliance.md, enrichment-playbook.md, data-sources.md). Each reference is one level deep, linked inline at the point of need with clear signals of what it contains (e.g., "for the verification actions, catch-all handling, and the suppression list"). This matches the 5 anchor (clear overview, well-signaled one-level-deep references); there is no nesting and no orphaned or buried content. | 5 / 5 |
Total | 18 / 20 Passed |