Conciseness

Dense and domain-specific throughout — the scope-filter table, byte heuristics, and report template all earn their tokens, and nothing explains concepts Claude already knows. It is not 5 because the `cd -- '<working_dir>' || { echo "BLOCKED..." }` guard block is repeated verbatim in three code snippets and "Read-only diagnostic ... No mutations" restates the description; not 3 because there is no real padding.

Actionability

Mostly executable: exact node CLI invocations with flags (`--tenant-id`, `--no-cache --require-tenant`), exact OData query strings for three of the five scopes, a numeric byte-per-row table, transfer formulas, and a complete report template. It is not 5 because the `WhoAmI` identity call is described in prose without its command line, and the team-owned scope filter is only approximated ("use team count from `teamroles_association`") with no concrete query; not 3 because the given commands are copy-paste ready, not pseudocode.

Workflow Clarity

The 5-step sequence is stated up front ("1. Verify project + locate profile → 2. Run verify → 3. Per-table row counts → 4. Cache-size estimate → 5. Report") and elaborated with explicit validation (drift check with "surface the drift list verbatim" and a reconcile recommendation), error-recovery paths (BLOCKED/NEEDS_CONTEXT exits, per-scope fallback "instead of blocking the whole preview"), and a terminal status-code contract. Not 4 because checkpoints and feedback loops are explicit rather than implicit.

Progressive Disclosure

Good structure for a self-contained skill: clear section headers, an upfront workflow summary, and well-signaled one-level-deep links to plugin-shared files (shared-instructions.md, app-working-directory.md); no bundle files exist to organize. It is not 5 because ~40 lines of estimate heuristics and the scope-filter mapping are inlined in a ~160-line file where a references/ file could carry them, and the sub-50-line exception does not apply; not 3 because nothing is buried and the inline content is core to execution.

Specificity

Names several concrete actions — "estimate the download size + sync frequency cost of an offline profile", "per-table row-count estimates", "Wraps verify-offline-profile.js" — all specific rather than generic. It stays at 4 rather than 5 because coverage centers on two related estimate actions rather than a comprehensive list of capabilities, and at 3 because it goes beyond 1-2 actions with fully concrete phrasing.

Completeness

Explicitly answers both: what ("estimate the download size + sync frequency cost of an offline profile", "Read-only", "Wraps verify-offline-profile.js with per-table row-count estimates") and when ("Use when the user wants to estimate ... BEFORE pushing to users") with a concrete trigger scenario. Not 4 because the "when" is explicit and scenario-specific, not merely present.

Trigger Term Quality

Good natural keyword coverage: "estimate", "download size", "sync frequency", "offline profile", "BEFORE pushing to users" — phrases a user would plausibly say. Not 5 because common synonyms are missing ("cache size", "data usage", "bandwidth", "preview"); not 3 because the terms present are natural and domain-relevant rather than generic.

Distinctiveness Conflict Risk

Clear niche — read-only pre-push cost estimation of an offline profile — with distinct trigger framing that separates it from sibling setup/edit/assign skills. Not 4 because the "Read-only" + "estimate ... BEFORE pushing" positioning makes overlap with mutating sibling skills minimal.