Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill body is highly actionable with clear workflows and concrete templates, but it is monolithic and verbose — detailed domain knowledge is inline with no progressive disclosure to reference files, and some prose could be tightened.
Suggestions
Split the per-mode 'Domain Knowledge' sections (Bill Preparation, LC Invoice Review, Client Billing Queries, Cashflow, Leverage) into separate reference files under references/ and link to them one level deep, leaving SKILL.md as a concise overview.
Trim editorializing prose (e.g., 'The write-down that avoids a difficult conversation today creates a realisation problem...') to imperative guidance that preserves the rule without the narrative.
Move the LC query-letter template and the three table schemas into a single references/output-formats.md to reduce inline bulk while keeping the BLUF-first and identifier-check rules in the main body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient specialized domain guidance Claude would not already know, but it includes editorializing prose ('The write-down that avoids a difficult conversation today creates a realisation problem across the matter that is harder to explain later') and narrative padding that could be tightened, so it does not reach the lean level-3 anchor. | 2 / 3 |
Actionability | It provides concrete, copy-paste-ready artifacts — the billing-instruction table schema, the LC query-letter template, the funding-gap and leverage tables, plus specific thresholds (5% write-down flag, £50k cashflow flag) — fully actionable instruction-level guidance. | 3 / 3 |
Workflow Clarity | Multi-step processes are clearly sequenced across five modes with explicit validation checkpoints ('Stop. Confirm identifiers before any output is produced', 'Confirm phase completion before billing', write-down partner authorisation), matching the level-3 anchor. | 3 / 3 |
Progressive Disclosure | The body is well-sectioned but monolithic at ~297 lines with no bundle files and no external references; detailed per-mode domain knowledge that could live in separate reference files is kept inline, fitting the level-2 'content that should be separate is inline' anchor rather than a one-level-deep reference structure. | 2 / 3 |
Total | 10 / 12 Passed |