Source-level design comparison · 19 October 2026 · two status codes
HTTP status lookup: reference, procedure or callable contract?
What each route adds
RFC 6585 §4 defines 429 as rate limiting and leaves user identification/request counting unspecified. RFC 9110 §15.6.4 describes 503 as a temporary overload or maintenance condition. Both allow Retry-After to indicate waiting; neither code alone supplies this customer's cause, counters or recovery time.
- Reference + ordinary prompt
- The original sections already supply the meanings. The reader chooses a source/version and a host that can read supplied excerpts or retrieve them. No lookup-specific endpoint or directory publication is inherent. No prompt answer was generated or compared.
- Proposed focused reference-backed skill
- A reusable source-reading/citation procedure could select the right section and distinguish missing current-state evidence. Its maintainer would own instructions and reference versions. It creates neither service access nor an enforced output contract. The skill is a design only: not authored, loaded or tested.
- Inspected remote-MCP tutorial
- Pinggy's HTTP-status example declares an integer input (100–599) for
lookup_http_status. The examined 429/503 entries come from a literal table; the handler returns text and structuredcode,phrase,meaning. No reference/provider fetch or customer-state input appears. The operator owns that callable contract, server, dependencies and table updates.
The shown Node server uses Streamable HTTP at /mcp; development routes client traffic through a tunnel. Reachable transport does not automatically refresh a literal table. Short summaries are not full references, but omission of optional retry detail alone establishes no defect. Schema/handler inspection does not prove runtime validation, invocation or result fidelity.
Choose the requirement before the public commitment
A named, maintainer-controlled callable contract is a legitimate reason to consider MCP even with static information. Learning or teaching its lifecycle is also legitimate. Neither establishes that this reader needs a publicly operated service for a meanings-only lookup, or that any route gives better answers.
Official Skills documentation distinguishes reusable instructions/resources from service capability. Public submission guidance adds stable HTTPS and publisher/domain/scanning/review commitments for the remote-public route, and separates hosted-tool updates from skills/metadata ZIP updates. A standalone skill, locally configured MCP connection and public plugin are different distribution units. An existing public skills-only plugin cannot currently gain MCP later; do not assume that in-place upgrade path.
One hypothetical changed requirement: “Why did this customer's actual request receive 429/503, and what is its current request/rate-limit/service state?” None of the inspected designs supplies that evidence. References explain meanings, the proposed skill guides reading, and the tool takes a number and reads its table. Appropriate actual request/provider/service evidence would still have to be supplied separately.
Source timing and untested boundaries
The relevant returned text was inspected within 19 October 2026, 12:52:57–12:54:10 UTC. This conservative bracket is not an individual fetch time, proof of fresh origin bytes or a source-change date. The tutorial is marked 1 October and its extraction says crawled last week. RFC/documentation extractions say crawled today; the official pages are undated. Only the cited component/public-commitment claims receive this check, not unrelated policies or host/catalogue studies.
The tutorial author claims 1 October validation, prints a local 429 response and reports a tunneled 503 lookup. Its ChatGPT call is prospective. These are author claims, not this comparison's observed results. No server, tunnel, dependency, endpoint, client/tool invocation, installation or public-submission step was operated. No prompt/tool output parity, reliability, cost, safety, demand or usefulness was measured; no actual customer state was obtained. The earlier studies retain their own dates, scopes and limitations.