Content
71%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.
The body is a lean, well-structured router with concrete commands, parameter-level critical rules, and real guardrail checkpoints. Weaknesses concentrate in scattered time-sensitive version data, one over-long rationale clause, and confusing MCP/sandbox conditional lines.
Suggestions
Consolidate the time-sensitive version data (the SDK version table, 'Latest Stripe API version: 2026-09-30.endive', and the '2026-03-25.dahlia' conditional) into a single clearly marked section or reference file so staleness is localized and the body stays tight.
Rewrite the three fragmented MCP/sandbox lines ('If stripe sandbox create is used, don't use MCP' / 'stripe sandbox claim' / 'stripe whoami') into one coherent rule stating exactly when to use MCP and what to run first.
Trim the rationale in the payment_method_types rule to its essential reason (dynamic payment methods, Dashboard-configurable) to reclaim roughly 20 tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and instruction-like (routing table, critical rules), but the time-sensitive items — 'Latest Stripe API version: 2026-09-30.endive', the 7-row SDK version table, and 'On API version 2026-03-25.dahlia or later' — sit outside any old-patterns/deprecated section, and the payment_method_types rule carries a padded rationale ('which enables you to configure payment method settings from the Dashboard and dynamically display the most relevant eligible payment methods to each customer to maximize conversion'). Anchor 3 ('mostly efficient but includes some unnecessary explanation or could be tightened') fits better than 4 because these instances recur. | 3 / 5 |
Actionability | Largely executable guidance: 'Install the Stripe CLI (`npm i -g @stripe/cli`) and run the command `stripe sandbox create`', 'Use `stripe whoami --format json` for CLI auth checks', 'pass the parameter `integration_identifier` to `checkout.sessions.create`', and exact key prefixes ('rk_' over 'sk_'). It falls short of 5 because the MCP/sandbox lines ('If `stripe sandbox create` is used, don't use MCP.' / 'first make sure `stripe sandbox claim` is run') are fragmented and leave Claude without a clear action. | 4 / 5 |
Workflow Clarity | A clear routing workflow exists — integration routing table, then 'Read the relevant reference file before answering any integration question or writing code', then critical rules — with a genuine validation checkpoint ('read the tax reference and confirm the user has an active registration' before enabling `automatic_tax`). Not 5 because there is no explicit ordering when multiple domains apply and the webhook rule defers its checkpoint to the reference. | 4 / 5 |
Progressive Disclosure | The ~60-line body is a true overview: a routing table pointing to all six real reference files (payments, connect, billing, tax, treasury, security — verified present and one level deep with no nested references), plus clearly signaled links like 'See security reference' and '[tax reference](references/tax.md)'. Content is appropriately split and navigation is task-indexed, matching the anchor-5 example. | 5 / 5 |
Total | 16 / 20 Passed |