preset-skill-op-endpoint-capabilities

Documents LLM preset endpoint declaration fields and Codex OAuth quota inspection procedures.

692|60|Updated Mar 15, 2026
One-click install
npx skills add https://github.com/Lingtai-AI/lingtai --skill preset-skill-op-endpoint-capabilities
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: preset-skill-op-endpoint-capabilities
Source: https://github.com/Lingtai-AI/lingtai/tree/main/tui/internal/preset/skills/lingtai-preset-skill/reference/operations/endpoint-capabilities
Command: npx skills add https://github.com/Lingtai-AI/lingtai --skill preset-skill-op-endpoint-capabilities

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It clarifies what each manifest.llm declaration field (provider, model, base_url, api_compat, capabilities) actually means versus credentials and live probes, preventing agents from conflating declaration-complete presets with working endpoints, and documents the exact Codex app-server operations for reading OAuth quota data.

Core Features & Use Cases

  • Declaration field reference: Defines provider, model, base_url, api_compat, and capability flags as non-credential declarations, distinct from credential validity checks in ResolveRefsWithAuth and from live /doctor probes.
  • Codex OAuth quota inspection: Gives the exact app-server sequence — initialize handshake, account/rateLimits/read request with null params, GetAccountRateLimitsResponse parsing, and optional account/rateLimits/updated subscription — plus the warning that codex doctor --json exposes no quota data.
  • Context-window evidence discipline: Separates the official 272K Codex context figure from a dated, route/model/CLI-qualified measured ~372K boundary, with rules for citing each.
  • Use Case: An agent asked "does this preset actually work?" routes the question to /doctor instead of guessing from manifest fields, and an agent needing Codex rate-limit data follows the documented JSON-RPC sequence without touching tokens.

Quick Start

Ask the agent to explain what the provider, base_url, and capabilities fields in a preset manifest declare and how to query Codex OAuth rate limits via the app-server.

Frequently Asked Questions about preset-skill-op-endpoint-capabilities

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What do the provider, model, and base_url fields in a preset manifest mean?

They are non-credential declarations: provider selects the kernel adapter class, model is the exact string sent to the provider, and base_url is the endpoint the adapter calls. They describe what will be called, not whether credentials are valid or the endpoint works.

How do I query Codex OAuth rate limits programmatically?

Complete the app-server initialize handshake first, then send an account/rateLimits/read request with null params. Parse the GetAccountRateLimitsResponse for rateLimits snapshots containing usedPercent, window duration, reset times, and credits fields.

Does codex doctor --json show quota or rate-limit information?

No. codex doctor --json emits a redacted installation, config, auth, and runtime health report with no rate-limit, quota, or usedPercent-shaped fields. Use the app-server account/rateLimits/read operation instead.

Why does a preset pass validation but still fail at runtime?

A preset can be declaration-complete and credential-valid yet fail a live probe due to provider outages, retired models, or exhausted quota. Save never tests declarations against real provider calls; live diagnosis belongs to /doctor.

What is the actual Codex context window size, 272K or 372K?

272K is the official vendor-published figure and the only number citable as current official. The ~372K figure is one dated synthetic probe measurement (OAuth route, gpt-5.6-sol, codex-cli 0.144.3, 2026-07-19) and must always carry those qualifiers.