Conciseness

Lean and dense: no explanation of concepts Claude already knows (no 'what is income tax' padding), every table row carries a rule plus where it bites in this specific codebase, and even the intro paragraph supplies only non-derivable project context (two trades, England-only). Matches anchor 5 — every token earns its place; anchor 4 would require sections that could be trimmed, and none are evident.

Actionability

As an instruction-only skill it gives fully actionable guidance: exact figures and dates for every rule (£37,700 / £125,140 bands, £7,105 threshold, £3.65/week, MTD thresholds by year), specific file paths ('src/utils/taxCalculator.ts', 'src/pdf/', 'references/rates-2026-27.md', 'evals/compliance/'), runnable commands ('npm run check', 'npm run eval:compliance'), and per-row statements of what is currently wrong in the code. This meets anchor 5 for an instruction-only skill per the scoring notes.

Workflow Clarity

'When you change tax code' gives a clear five-step sequence with explicit validation checkpoints (write the compliance case first with source and worked example, then run 'npm run check (regression gate) and npm run eval:compliance', then promote passing cases into 'tests/'). It falls between anchors 4 and 5: validation steps are explicit (anchor 5) but no error-recovery feedback loop ('if the gate fails, fix and re-run') is spelled out, matching anchor 4's 'minor validation gaps'.

Progressive Disclosure

A clear overview that centralizes all figures in one well-signaled, one-level-deep reference that actually exists and is well-sourced ('Every figure lives in references/rates-2026-27.md with its gov.uk source'; step 1 'Read references/rates-2026-27.md'; verified the file exists with gov.uk URLs and per-year tables). The SKILL.md body carries only rules and app-specific context, which is the appropriate split. Matches anchor 5.

Specificity

The description enumerates ten concrete rule areas ('income tax bands, personal allowance taper, Class 2/4 NIC, cash basis, payments on account, more than one trade, trading allowance, Making Tax Digital, invoice content') with no fluff, but capabilities are stated as topic nouns rather than actions — what the skill does with these rules is only implied by 'tax rules for this invoicing app'. This is above anchor 3 (which covers only 1-2 concrete items) but falls short of anchor 5's 'multiple specific concrete actions'.

Completeness

Explicitly answers both questions: the 'what' is 'UK self-employed (sole trader) tax rules for this invoicing app' with an enumerated scope, and the 'when' is concrete — 'Use before changing anything under src/utils/taxCalculator.ts, the Dashboard or Accounting pages, expense handling, or PDF invoice content, and when writing or reviewing tax tests.' This matches anchor 5 exactly; anchor 4 would require a weaker or less specific 'when'.

Trigger Term Quality

Strong natural terms a user would actually say: 'UK', 'self-employed (sole trader)', 'tax', 'income tax', 'NIC', 'cash basis', 'payments on account', 'invoice'. Anchor 5 is not met because common synonyms are missing — 'Self Assessment' (the return UK sole traders file), 'National Insurance' spelled out, and 'HMRC' are absent, leaving only the jargon 'Class 2/4 NIC'.

Distinctiveness Conflict Risk

A clear niche — UK sole-trader tax for one specific app — with triggers pinned to concrete file paths ('src/utils/taxCalculator.ts', 'PDF invoice content', 'tax tests'), making accidental triggering by other skills (e.g. general invoicing or PDF skills) very unlikely. Matches anchor 5's 'clear niche with distinct triggers; minimal conflict risk'.