Conciseness

Every line is a rule, example, or exception — "Respond terse like smart caveman. All technical substance stay. Only fluff die." There is no padding and no explanation of concepts Claude already knows; the body even models the compression style it teaches. It matches the lean, every-token-earns-its-place anchor.

Actionability

Guidance is fully concrete and executable: explicit drop-lists with example words ("just/really/basically/actually/simply"), specific abbreviations (DB/auth/config/req/res/fn/impl), a response pattern template ("[thing] [action] [reason]. [next step]."), a bad-vs-good contrast, and two worked examples covering common cases. This is copy-paste-ready instruction for a behavior skill.

Workflow Clarity

This is a simple, single-purpose response-style skill and the single action is unambiguous. The persistence rules ("ACTIVE EVERY RESPONSE once triggered... Off only when user says 'stop caveman'") and the Auto-Clarity Exception with an explicit resume step provide clear checkpoints, satisfying the simple-skill exception.

Progressive Disclosure

The skill is under 50 lines, self-contained with no bundle files, and organized into clear sections (Persistence, Rules, Examples, Auto-Clarity Exception). Per the rubric's simple-skill guidance, well-organized sections alone warrant the top score when no external references are needed.

Specificity

The description enumerates multiple concrete, measurable behaviors — "Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy" — with no vague filler language. It matches the anchor for multiple specific concrete actions with comprehensive coverage; nothing material is missing for this skill's domain.

Completeness

It explicitly answers both what the skill does (ultra-compressed communication preserving technical accuracy, with the specific mechanisms) and when to use it, with concrete trigger phrases in a "Use when..." clause. This matches the anchor-5 example structure exactly.

Trigger Term Quality

Six natural trigger phrases plus the slash invocation are listed: "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. This comprehensively covers the natural terms and variations a user would actually say, matching the top anchor rather than merely having good coverage.

Distinctiveness Conflict Risk

The caveman-specific triggers ("caveman mode", "talk like caveman", "/caveman") form a clear niche, but the generic phrases "be brief" and "less tokens" could plausibly trigger other brevity or token-reduction skills — minor overlap risk with closely related skills, so it sits at anchor 4 rather than 5.