Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A concise, well-structured verification workflow with explicit validation checkpoints appropriate to a read-only reconciliation skill. The main gap is actionability: the computation is fully specified but the actual NAV data retrieval is left abstract without naming the MCP tools or calls to use.
Suggestions
Name the specific nav MCP tools/calls (or the exact pack fields) used to retrieve LP commitment %, fund-level P&L components, fee/expense totals, and waterfall outputs so the retrieval step is executable.
Add one concrete worked example showing a recomputed line alongside the statement value and the resulting flag (e.g., the 12.40% vs 12.38% ownership case) to make the compare-and-flag step copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean body with no padding or explanation of concepts Claude already knows; the recompute formula, tolerance, checks, and output spec each earn their place. | 3 / 3 |
Actionability | The recomputation formula, 0.01 tolerance, and per-line comparison instructions are concrete, but the data-retrieval mechanism is underspecified — 'via the nav MCP' and 'Pull each input from the NAV pack' name no specific MCP tools or calls, leaving the execution path incomplete. | 2 / 3 |
Workflow Clarity | A clear Recompute → Compare → Additional checks → Output sequence with explicit validation checkpoints (per-line comparison at 0.01 tolerance plus the cross-footing checks); the skill is read-only by design so no fix-retry loop is required. | 3 / 3 |
Progressive Disclosure | Under 50 lines, single-purpose, with no need for external references; the well-organized sections satisfy the simple-skill exception for the top level. | 3 / 3 |
Total | 11 / 12 Passed |