Content
77%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-engineered skill body: explicit workflow with real validation gates, concrete commands and thresholds, and exemplary progressive disclosure verified against the actual bundle. The main weakness is token efficiency — the 'When to Use' section and inline analysis bullets duplicate content available in the frontmatter and reference files, and a few passages could be tightened.
Suggestions
Remove or compress the 'When to Use' example-query list, which restates the frontmatter description's trigger clause, and fold any novel triggers back into the description.
Trim the Step 3 inline dimension bullets (asset class, sector, market cap, geography) to just the decision-relevant thresholds, leaving framework detail to the already-referenced files like references/asset-allocation.md.
Make Step 2's enrichment guidance concrete by specifying which MCP market-data tools or REST endpoints to call, mirroring how Step 1 defers to references/alpaca-data-fetch.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — no basic-concept explanations, and analysis bullets carry domain thresholds Claude would not reliably know (e.g., 'Optimal range: 15-30 stocks', 'flag if any sector exceeds 30-40%') — but there is tightening to do: the 'When to Use' quote list duplicates the frontmatter triggers, the Step 3 dimension bullets partially restate content already delegated to reference files, and the 'Limited Offline CSV Review' paragraph is dense and meandering. It sits above the 'noticeably verbose' anchor 2 but clearly below 'efficient with only minor trims'. | 3 / 5 |
Actionability | Most guidance is executable: copy-paste bash with env vars for the connection check, exact script paths and flags (`scripts/reconcile_positions.py`, `--input-format alpaca-json`, `--confirm-complete`), a concrete report filename format, and named MCP tools. It misses the 5 anchor because Step 2 enrichment relies on vague direction ('Use WebSearch or available market data APIs') and report generation has no worked example, leaving minor gaps. | 4 / 5 |
Workflow Clarity | The 7-step sequence is explicit with sub-steps and genuine validation checkpoints: data-validation bullets in Step 1, a hard reconciliation gate ('Stop allocation and rebalance review if can_continue is false'), a connection script whose exit code only signals success when both endpoints succeed, and an Error Handling section with recovery paths for four failure modes. This matches the anchor for clear sequencing, explicit validation, and error-recovery feedback loops. | 5 / 5 |
Progressive Disclosure | The body is an overview that points to 15 reference files, 4 scripts, and 3 assets, each referenced by exact path and named section (verified to exist), references are one level deep with no cross-nesting, and a dedicated `references/resource-index.md` provides per-file When/Contains navigation. This matches the anchor for a clear overview with well-signaled one-level-deep references and easy navigation. | 5 / 5 |
Total | 17 / 20 Passed |