Conciseness

The body is dense procedural guidance that assumes Claude's competence and does not explain basic concepts, with concrete parameters ('format: "none"', 'entity_type: "person"'); it earns 4 rather than 5 because some emphases repeat across sections ('size is an output' appears in Purpose, step 1, How to behave, and Refuse cleanly) and a few long parentheticals could be trimmed.

Actionability

It gives mostly executable guidance — a concrete command ('cat "<Bundle root>/context/scoring.md"'), specific parameters, role bounds ('5–20 defining, 0–3 must-haves, 0–5 exclusions'), and exact build semantics ('defining any-of, must-haves all-of, exclusions none-of'); it stops short of anchor 5 because the detailed procedures are deferred to referenced context files rather than fully copy-paste-ready inline.

Workflow Clarity

The flow is clearly sequenced into four numbered phases with checkpoints ('let the operator amend before any search', 'end the turn at the pivot question', 'Exclusions are explicit-include only') plus a Failure modes section with recovery guidance; it is not 5 because the main flow lacks an explicit validate→fix→retry loop of the kind anchor 5 illustrates, though the operation is read-only so the destructive-cap does not apply.

Progressive Disclosure

Structure is good — a clear overview (Purpose, Works with, Language) with one-level-deep, well-signaled references to context files ('context/discovery.md', 'context/scoring.md', 'context/profiling.md'); it does not reach 5 because no bundle files exist in references/scripts/assets to confirm the referenced paths resolve, and references use a '<Bundle root>' placeholder rather than verified local paths.

Specificity

It names the domain ('Read who a market is') and three process verbs ('discover → pivot → read'), but these are high-level procedural labels rather than concrete capabilities like 'extract text' or 'fill forms', so coverage is not comprehensive — matching anchor 3 and falling short of the multi-action specificity of anchor 4.

Completeness

It explicitly states both what ('Read who a market is... discover → pivot → read... Aggregates only — never individual records, never an export') and when ('Use when a read-shaped ask arrives with a brief and no signals yet') with concrete trigger phrases, matching the anchor 5 example; it is not anchor 4 because the 'when' is fully explicit rather than merely adequate.

Trigger Term Quality

It quotes three natural phrases a user would actually say ('"who's in the market for X"', '"profile this audience"', '"an audience profile for my client"'), giving good keyword coverage; it stops short of anchor 5's comprehensive synonym/extension coverage since the triggers are narrow to read-shaped audience asks.

Distinctiveness Conflict Risk

It carves a clear niche ('The discovery-first way into audience-analyze, behind /watt:audience', 'Not a user command') and explicitly separates from siblings ('size is an output, never a target' vs generate; 'never an export' vs activate), giving minimal conflict risk as in anchor 5.