Conciseness

The body is extremely lean: every rule is one compact line ('One particle, one relation.', 'Identifiers verbatim, original script? Negations present? Two readings? 白話.'), and nothing explains concepts Claude already knows — even the Classical-Chinese grammar notes are stated as terse rules, not lessons. Matches anchor 5 ('lean and efficient; assumes Claude's competence; every token earns its place'); anchor 4 would require trimming opportunities that don't exist here.

Actionability

Concrete, executable guidance throughout: four Bad/Good example pairs covering the common response cases (rendering explanation, connection pooling, code-change summary, payload terminology), exact commands ('/caveman status', 'stop caveman', 'normal mode'), fully specified status-reporting behavior ('relay the hook's Caveman mode: <mode> value; with no hook, the mode you were in (or off) plus (not tracked by this host)'), and a pre-send checklist. For an instruction-only skill this matches anchor 5's 'specific examples cover the common cases'.

Workflow Clarity

This is a simple, single-purpose register skill with no multi-step or destructive/batch operations, so the simple-skill exception applies: the single action (respond in 文言) is unambiguous, persistence has an explicit default ('Unsure? Still on.'), and the 'Pre-send check' section is an explicit validation checkpoint before every response, plus clearly enumerated break conditions in 'When to break the rules'. This satisfies anchor 5's clarity bar; anchor 4 would imply missing checkpoints, but the pre-send check covers that.

Progressive Disclosure

No bundle files (references/, scripts/, assets/) exist and none are needed at this size: the body is a short, well-organized set of sections (Persistence, Floor, six numbered rules, break conditions, pre-send check) with no nested references and nothing that belongs in a separate file inlined. Under the rubric's simple-skill note, well-organized sections alone earn anchor 5 here.

Specificity

The description states several concrete behaviors — '文言文 register, far fewer characters, technical terms verbatim', persistence semantics ('Stays on until "stop caveman" or "normal mode"'), and status reporting via '/caveman status' is not needed here but the register/character/payload behaviors are each specific. It falls just short of anchor 5 because the actions are stated tersely rather than comprehensively (e.g., no mention of fallback to 白話 or break conditions), matching anchor 4's 'several specific actions; minor gaps in coverage'.

Completeness

What: 'Caveman in Classical Chinese: 文言文 register, far fewer characters, technical terms verbatim.' When: 'Invoke only with /megacave or /caveman wenyan', plus explicit end conditions ('Stays on until "stop caveman" or "normal mode"'). Both what and when are explicitly answered with concrete trigger phrases, matching anchor 5; anchor 4 would require the 'when' to be less explicit than the direct 'Invoke only with' clause.

Trigger Term Quality

Triggers are explicit and natural for this need: '/megacave', '/caveman wenyan', 'caveman', 'Classical Chinese', '文言文', plus off-phrases 'stop caveman' and 'normal mode'. Coverage is good but misses common synonyms like 'literary Chinese', '古文', or plain 'wenyan' as a standalone term, fitting anchor 4 ('good keyword coverage; a few natural terms missing') rather than anchor 5's comprehensive synonym/extension coverage.

Distinctiveness Conflict Risk

'Caveman in Classical Chinese: 文言文 register' is a clear niche with virtually no overlap with other skills, and 'Invoke only with /megacave or /caveman wenyan' removes any auto-trigger conflict risk. Matches anchor 5 ('clear niche with distinct triggers; minimal conflict risk'); it is not anchor 4 because there is no plausible overlap with a closely related skill given the explicit-only invocation.