spb/llm-api-toolkits

1# claude-api-toolkit23An Agent Skill for writing, reviewing, debugging and migrating code that calls the Claude API.45What makes it different: its model tables and its request linter were built by sending 46 request6shapes to each of 15 Claude models and recording what the API accepted (see7`references/request-rules.md`). The linter's verdict matched the API on every recorded outcome.89Not affiliated with Anthropic. "Claude" is a trademark of Anthropic, PBC.1011## Contents1213| Path | Purpose |14|---|---|15| `SKILL.md` | When to use the skill, the working loop, the rules that cause most failures |16| `references/models.md` | Ids, context, output caps, prices (generated) |17| `references/request-rules.md` | What each model accepts, with exact error texts (generated) |18| `references/requests.md` | Request anatomy, streaming, structured outputs, files, batches, token counting |19| `references/tools-and-agents.md` | Tool definitions, the tool loop, server tools, agent architectures |20| `references/managed-agents.md` | Core Managed Agents flow |21| `references/caching-and-context.md` | Prompt caching, compaction, context editing, cost levers |22| `references/errors.md` | Status codes and observed 400 messages with fixes |23| `scripts/claude-api.mjs` | Offline CLI: `models`, `lint`, `fix`, `migrate`, `cost`, `compare`, `scaffold`, `guide` |2425The scripts use only the Node.js standard library, open no network connection and need no key.2627## Install2829```bash30skills install @spboucher/claude-api-toolkit --client claude-code   # or codex, cursor, generic-agent-skills31```3233## Measured effect3435`tests/skill-eval` in the source repository asks a model to write Messages requests for specific36models, with and without this skill in its system prompt, and sends each result to the live API.37The measured acceptance rates are in `CHANGELOG.md`. It is a small evaluation (see the numbers and38their caveats there), not a benchmark.3940## Limits4142A snapshot taken on the date in `SKILL.md` under API version `2026-10-21`. Models the probing key43could not see are absent. Computer use, the advisor tool, memory stores, message threads and44cloud-platform differences were not probed. Cost output is an estimate. When the API and this45skill disagree, the API is right.46