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 dense, highly actionable platform reference: real limits, dates, flags, and copy-paste code for every feature, plus excellent diagnostic decision trees with explicit error-recovery loops. Its weaknesses are repetition — the anti-Edge guidance and key numbers are restated across five or more sections — and the complete absence of progressive disclosure, with ~725 lines all inlined in SKILL.md and no reference files.
Suggestions
Consolidate the Edge-runtime guidance: it is argued in "Prefer Node.js over the Edge runtime", then repeated in the Choosing-a-Runtime note, the Streaming callout, Common Pitfalls #3, and both the cold-start and edge-timeout diagnostics trees — state it once in the main section and cross-link from the others using the existing anchor links.
Split the bulk sections into one-level-deep reference files (e.g. references/containers.md, references/websockets.md, references/diagnostics.md), leaving short summaries plus explicit pointers in SKILL.md so the body works as an overview rather than a monolith.
Deduplicate repeated numbers: the 300s/800s/1800s durations and the 1/2/4 MB Edge code sizes each appear in 3–5 tables and sections — state them once in "Plan Limits at a Glance" and reference that section elsewhere.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Nearly every line carries Vercel-specific facts Claude won't know ("enabled by default for new projects since April 23, 2025", "Next.js 16.3", "VERCEL_SUPPORT_LARGE_FUNCTIONS"), so there is no padding with known concepts — but the ~725-line body restates the anti-Edge message in five-plus places ("Strongly prefer Node.js", "You do NOT need runtime = 'edge' for streaming", pitfall #3, both the cold-start and edge-timeout trees) and repeats the 300s/800s/1800s and 1/2/4 MB numbers across multiple tables. Mostly efficient, but it could be tightened by 20–30%, which fits anchor 3 better than 4 (the repetition goes beyond minor) and far above 2 (nothing explains concepts Claude already knows). | 3 / 5 |
Actionability | Copy-paste-ready throughout: `export const maxDuration = 800`, the waitUntil/Promise snippet, the SSE ReadableStream handler, complete `ws` and Socket.IO servers with the required client config (`transports: ['websocket']`), a working Dockerfile.vercel + srvx pair, `vercel env add VERCEL_SUPPORT_LARGE_FUNCTIONS`, `vercel vcr` commands, and cron config with a CRON_SECRET verification handler. Concrete error codes (`504 FUNCTION_INVOCATION_TIMEOUT`, `413 FUNCTION_PAYLOAD_TOO_LARGE`) each map to specific fixes — a full match for the executable anchor. | 5 / 5 |
Workflow Clarity | Multi-step processes are clearly sequenced with checkpoints: the Edge migration is a numbered 4-step list ending in "Confirm Fluid Compute is on (default since April 23, 2025) and redeploy", and the four diagnostics trees (timeout, 500, invocation failure, cold start) are explicit error-recovery feedback loops branching symptom → check → fix. The duration tier guide ("≤ 300s → any plan... beyond 30 min → Vercel Workflow") acts as a checklist for the most common decision. | 5 / 5 |
Progressive Disclosure | The body has clear section headers, a plan-limits summary, and in-document anchor links, but there are no bundle files at all — everything is inline in one ~725-line SKILL.md. Sections like Docker and Container Images (~75 lines), WebSockets (~95 lines), and the diagnostics trees are exactly the content the rubric wants in one-level-deep reference files. That matches anchor 3 ("Some structure but... content that should be separate is inline") — better organized than anchor 2's header-less wall, but short of anchor 4's "most content is appropriately placed". | 3 / 5 |
Total | 16 / 20 Passed |