The description names the domain (referral/affiliate programs) and mentions actions like 'create, optimize, or analyze,' but doesn't list specific concrete deliverables or outputs (e.g., 'design reward tiers, calculate payout structures, draft referral emails').

Clearly answers both 'what' (create, optimize, or analyze referral/affiliate/word-of-mouth strategies) and 'when' (explicit 'Use when' clause with trigger terms, plus a clarifying boundary with the launch-strategy skill).

Excellent coverage of natural trigger terms including 'referral,' 'affiliate,' 'ambassador,' 'word of mouth,' 'viral loop,' 'refer a friend,' 'partner program,' 'referral incentive,' 'how to get referrals,' 'customers referring customers,' and 'affiliate payout.' These are terms users would naturally use.

The description carves out a clear niche around referral and affiliate programs, and explicitly differentiates from a related skill ('For launch-specific virality, see launch-strategy'), reducing conflict risk significantly.

The skill is reasonably well-organized but includes some unnecessary explanatory content that Claude already knows (e.g., explaining what referral vs affiliate programs are, typical characteristics). The 'Referral vs. Affiliate' section and some of the context-gathering questions add bulk without adding unique instructional value. The email template examples and statistics ('16-25% higher LTV') are useful but the overall document could be tightened.

The skill provides structured checklists, a problem/fix table, and email templates which are somewhat actionable. However, it lacks concrete executable code, specific implementation steps, or copy-paste ready configurations. Much of the guidance remains at a strategic/advisory level ('simplify sharing process,' 'test different incentive types') rather than providing specific, concrete instructions Claude can directly execute.

The launch checklist provides a clear sequence of steps, and the referral loop (Trigger → Share → Convert → Reward → Loop) gives a good conceptual framework. However, there are no validation checkpoints or feedback loops for error recovery. The optimization section lists what to do but doesn't sequence it into a clear diagnostic workflow with decision points.

The skill effectively uses progressive disclosure with clear references to external files: program examples in references/program-examples.md, affiliate details in references/affiliate-programs.md, and tool-specific guides linked in the tools table. References are one level deep and clearly signaled. The main document serves as a well-organized overview with appropriate depth.