Claude API

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites — Paid Anthropic API usage is required to USE what this skill helps build, and that cost is mine: the skill itself is a documentation reference (installing its files needs no account, no key and no payment, and calls no API), but the code written with it calls the Anthropic API with my own authentication and is billed by Anthropic per token used. Before writing or running such code, use my existing authentication first (the SDK and ant CLI credential chain, an ant auth login profile, or ANTHROPIC_API_KEY if I have set one); do not ask me to paste a key unless ant auth status shows no active credential source. Some workflows in the skill (evals, cost optimization, hill-climbing, replays) spend real API money on every run and need my approval of the budget first; organization admin features need a separate admin credential. Installing the files does none of that and sets up no credential. Install "Claude API" for me. It is a large reference the agent consults before writing code that uses Claude: model ids, pricing and limits, request parameters, streaming, tool use, Managed Agents, prompt caching, token counting, error handling, eval building and model migration, organized as one main SKILL.md plus per-language folders (Python, TypeScript, Go, Java, Ruby, PHP, C#, cURL) and shared topical files. It tells the agent to write code through the official Anthropic SDK for the project language (raw HTTP only when asked or when no SDK exists), to never guess SDK usage from memory, to check a target file for another provider first and stop rather than rewrite it, and to resolve credentials from the existing authentication chain instead of asking for a key. Subcommands cover migrating a model, auditing prompts, upgrading the Python SDK, optimizing cost and building or hill-climbing evals, several of which spend API money per run and ask for budget approval first. Repository: https://github.com/anthropics/skills (the skill is the whole skills/claude-api/ folder at the pinned commit: SKILL.md plus 95 companion files, 96 files in total) Pinned version: install the code at commit dbd4588f9e1033efb41dad4bef2f7947c8993d44 exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/anthropics/skills into a temporary folder, check out commit dbd4588f9e1033efb41dad4bef2f7947c8993d44, and copy the skills/claude-api/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-22: [FINGERPRINTS] - LICENSE.txt: SHA-256 bc6b3af2f331cbc7fb0da1344efb2cbe5877a31498b4d70dbc7000f3405a1362 - SKILL.md: SHA-256 40537cca17c4aa86e4ce83ced39efcbfb1204905e3cf4c4ff98f2eab011e4b7f - csharp/claude-api/README.md: SHA-256 87a2cc5c6e670e0bf34120f9a592d321567eea57c716c06791f6d2dcaba08b00 - csharp/claude-api/batches.md: SHA-256 ec276fccf21757ed00803dfa4bc35f14f334138629fa53c77a025d718eda96a1 - csharp/claude-api/files-api.md: SHA-256 766ccf5dca9ef3caba8cefeddde10f6d171a7bd2ff4c97689462fad2d0d18e4b - csharp/claude-api/streaming.md: SHA-256 bc0100e50b3495e869901bcfb231870cb3483b936534af5fd3dddb5d07e4531e - csharp/claude-api/tool-use.md: SHA-256 18aa857c41e39eedb35815bfd8d50868a1bbc1f1a657f5e7233a3879baebc764 - curl/examples.md: SHA-256 4520e447d37924e62fe982923f3eceea5338dbf3b3a0d25d0f914272c16b90a0 - curl/managed-agents.md: SHA-256 faa7dd3d83cfd32d1faabb13fee5902e041285579b00256f17fd21681ed3ff2a - go/claude-api/README.md: SHA-256 c73ac644d8efe65a8d149e800267f98b0aa9d1e716ca38a22fd8803d5c35bfae - go/claude-api/files-api.md: SHA-256 7bf67ece5e4d18b1fb07085a55b680420cc7bd4de94a1fb2bb73cb23f39c20f0 - go/claude-api/streaming.md: SHA-256 a523e992e438309ca1bbc075ded2b652eb257242fd0ce521afd3b8ea024ed94d - go/claude-api/tool-use.md: SHA-256 18c6dc0cec151293b44697c340e0e0759f6081ad711ca34b06570c2450f243e9 - go/managed-agents/README.md: SHA-256 0343be8d49f94687fe5e068ad69d9e05253b408e7bcd0e230b39924f6404facc - java/claude-api/README.md: SHA-256 cb6d395dc1d930371f828290132681541eff7f1a7e2739bb4c6655ebbb3b8c7f - java/claude-api/files-api.md: SHA-256 1c792f3421de32f42a56235327f87807a5579340726bba677ec13e05f9b493c7 - java/claude-api/streaming.md: SHA-256 de13bc9573b37edb5af69606853c4fde3d494ee10ff47d1a030ecef9cc17463b - java/claude-api/tool-use.md: SHA-256 1916b8d4e49b5d319a2d0c7e21b5a3d7a0f225e4a94d82139e769f66eb66f6d1 - java/managed-agents/README.md: SHA-256 ba2a845898b302dc78b5354aa8f34ae4fc2c2ea41d12aa7d0440db59ee004e0c - php/claude-api/README.md: SHA-256 00e2e88f3ea9a6cce812d3e6ae9e035bc36f88828a09e518bb73af91097a9d66 - php/claude-api/batches.md: SHA-256 6f82a29f718ff6e6b02aaaee3cfd6cda2feb82aca57d945f11a2c021214a699d - php/claude-api/files-api.md: SHA-256 0715a0d4cce88fd885f57bd33b4654071a9068d5f8a892143b062c78fe51afc9 - php/claude-api/streaming.md: SHA-256 3b8ac667b21f17750eb3c1fc7dc166fd7b1f57a1d8b11e74489de4432e21d31d - php/claude-api/tool-use.md: SHA-256 1d79c316ef608d8327da1ae7495ecc4a0c517bdd761037c886d8fc61a0eca7d0 - php/managed-agents/README.md: SHA-256 2f9ada8a85098a7a4a7265c29fadb74d695a33bf19bc6d5df3a2c275b4a535b2 - python/claude-api/README.md: SHA-256 284fbc6408809afd27288da95c5fb32daa48bd77518cd5fd4e973cc05363eaa8 - python/claude-api/batches.md: SHA-256 7e4d07173845836617be3db70482c182311ad2d6d5ddf786482ca57c7c45776c - python/claude-api/files-api.md: SHA-256 2aa142978660d9ae3200d21a4d333531ca101898706f620834304e8a7d534580 - python/claude-api/sdk-upgrade.md: SHA-256 f7c5651b21dd6fd5651935415d907fa0a369736f3055d50edb3d236ea016527c - python/claude-api/streaming.md: SHA-256 8f73a54dbbb5e82c2c7690d7d866f17ad8c72f662a2be6dd788b5c0956714026 - python/claude-api/tool-use.md: SHA-256 548dcc4bd2a36e8b90cec446f9b0b7917d2d75a085c7c2603b28c840bd621621 - python/managed-agents/README.md: SHA-256 82b6e5847027b0ffe5976116e5c249a58e9e2781eb09e914a81ee2f9eb970245 - ruby/claude-api/README.md: SHA-256 c7de89d236f22bc0b44a091a45ee186ec057caa021a71acd4d3849599cbf3b36 - ruby/claude-api/streaming.md: SHA-256 09b784f2ab0a5fd17e0589225408c618d26c6dd3cd601364ff92f409d5b104a0 - ruby/claude-api/tool-use.md: SHA-256 80cbb7c7fd4a01b816524eb54e0f8f4e5b4cc7643d393dfce6b9682e0f4d61b6 - ruby/managed-agents/README.md: SHA-256 729ba22b8711a09192006d169098f16b933dd124ff36421772e4334d862306ab - shared/admin-api.md: SHA-256 a69c53817391fca50c0a1848f46de9debf71fa3681298fccb0a6435b760bac2d - shared/agent-design.md: SHA-256 9960728faf6976a2ddfceed02ed72af122d3b2558af573a9a0ea207b3adec2bc - shared/anthropic-cli.md: SHA-256 2baa725ea0248ab7d98dcb218f09e5eb22f768006ad28f4d5011b4975b4d500e - shared/claude-platform-on-aws.md: SHA-256 dec03193c61909167846c0526a4feacf011578b5e7ee29f76378141315cf97f5 - shared/cost-optimization.md: SHA-256 d0107a954e9059f43c0aaac07ac1f8a9f5aa0a898834d3d983d623e5976e051b - shared/error-codes.md: SHA-256 1abf8dcc262732f7ee5e4c58d1909fbb0722ac12e1b957bffdf64f28fde9e4d2 - shared/evals/build-eval.md: SHA-256 c8f6ae960867f6b3b35557aec75361553888ae7168f268365444bf6a4e09a5d7 - shared/evals/cost-hillclimb.md: SHA-256 dd46a354fed24d96c7f9acf42f9da64f5cceaa797949662d82374fd2cf7a3515 - shared/evals/eval-audit.md: SHA-256 a013f8eb0452c51c5a31860df0fb47219dcdb4da507092fc98867225095026ba - shared/evals/eval-hillclimb.md: SHA-256 315ae8540ebaeb9f1d7de5efa49625e6137ead255dc3a545b63393df89375588 - shared/evals/report/SCHEMA.md: SHA-256 9a83404a9b3c413113df39d9bdf08e003ff4d6a75386f4010887da3298766687 - shared/evals/report/build-report-lite.mjs: SHA-256 af79ca5fa5f06747dbcaa1d40017c8c8e81e0aa11cb8e32b848043893cdb03c7 - shared/evals/report/runner-scaffold.mjs: SHA-256 c3a87ba4b94b3018b83c075d5148ba7760580aa447699d513d45ab55b82a531a - shared/live-sources.md: SHA-256 cc82c88587c861919557598df0421ea7e8cec683f63adead997a0170dc78732b - shared/managed-agents-api-reference.md: SHA-256 efd84dc8a17a2842965c775a905b17fa94cf762cc544df81e708f3aa7a89ae7b - shared/managed-agents-client-patterns.md: SHA-256 e9cf54ca98a9cdd7dba0f51acfedfc1b0f160bad243e39f338ff39d4bf5bf823 - shared/managed-agents-core.md: SHA-256 b1e7f90f644ef3333bc2c8407ba17410cf672314868474c73242ac969e22007f - shared/managed-agents-environments.md: SHA-256 227e52e6a3d1c967c68912738ff184a97d1f4cd019c6a7bc7cd90c9b9857e5a5 - shared/managed-agents-events.md: SHA-256 3c5edc1428bc4b9cff1ebf01cecffaff3ef1650c19afa48c616fcd05b9586e03 - shared/managed-agents-memory.md: SHA-256 0dec5641630f97629ea75bbbe5c2bced861f9c16b0572a68c901849ea4d00bcb - shared/managed-agents-multiagent.md: SHA-256 f3d03cc910064cf419c80ce79c5fe62abde127467deea55491b447a07953a06b - shared/managed-agents-onboarding-from-quickstart.md: SHA-256 e37628c5639786a7868acfed4be0e25678b69b7d13a9aced03d74deebe042342 - shared/managed-agents-onboarding-from-url.md: SHA-256 b6c49289186d7cc521c995ba88fa916b65433101b2b3e9bf3c9add94ad87fa34 - shared/managed-agents-onboarding.md: SHA-256 025683a9ee3db081d8a6978980349ce68d1df316d0c5aea31ae12438907208c9 - shared/managed-agents-outcomes.md: SHA-256 a516c6bf6a6a76f7f5ef83ec80895efd5de718e8f9418b57c705ce151afc5afe - shared/managed-agents-overview.md: SHA-256 abcbdc29435c07be4749e21f005ed51f45942af03fa0bdcdeb18ebfa38c6a6f8 - shared/managed-agents-quickstarts/bug-hunter.md: SHA-256 9afd40ef3e1bca9e1cc4c6655d314127cec0b4dcff0e12f6016508c134954248 - shared/managed-agents-quickstarts/contract-tracker.md: SHA-256 312ee5f886de59b6db4603d2df9766d66285db36973656c5b9ff0bcc1ed02279 - shared/managed-agents-quickstarts/data-analyst.md: SHA-256 c2116bac43972f2b40311104bd7cd4e084cca806dd40a1ddf8462845d785bd57 - shared/managed-agents-quickstarts/deep-researcher.md: SHA-256 a680856d874dff23cd3f016e36d0a58bc4db36c56d790d216743f236ee37a48c - shared/managed-agents-quickstarts/explainer-video-maker.md: SHA-256 c2570c5e14d51c5caa77646ac417d18619d34a27cc94dc72cf020c698f2803f3 - shared/managed-agents-quickstarts/field-monitor.md: SHA-256 f17f26eee0109f4172134c5bea5761c52b8459e44ae03608c701cd3492f4a3af - shared/managed-agents-quickstarts/incident-commander.md: SHA-256 badd3622ffab0798ab44a55189f47e95c107107ce57eda9fa5357b1aded8fb2a - shared/managed-agents-quickstarts/performance-tuner.md: SHA-256 922396e14b569bc98a7bdde34d10890cc81c5184e81971e6c9be108d191dd639 - shared/managed-agents-quickstarts/sprint-retro-facilitator.md: SHA-256 8b5ed9895feb1ade54578a270493c63d0f57b3bb0fbc2eafcfa2e87d3130fa75 - shared/managed-agents-quickstarts/structured-extractor.md: SHA-256 c278074c8c48d35fc6c68ed32eaa8704cae73a8e74035fd1d4a6abdcb9bd0fbd - shared/managed-agents-quickstarts/support-agent.md: SHA-256 da75c50edf207f5ef468fbbbfc2e77d56e7d906a6c2cda28756f896195e171c5 - shared/managed-agents-quickstarts/support-to-eng-escalator.md: SHA-256 46720e5341808a9938825be44f936c91f83739ebdef9e1d17d9f0ec22738346b - shared/managed-agents-quickstarts/watchlist-scanner.md: SHA-256 e67861c63b9d46b01c8d2784998b7f0de9a87aa6012b77127def1253c5ac13db - shared/managed-agents-scheduled-deployments.md: SHA-256 3fdf7bd691b4f798149d7ebdb933dcee7d14d5d47dd6bf8196c41b819223175e - shared/managed-agents-self-hosted-sandboxes.md: SHA-256 a0c0d98ee1921ee5856a9fd1c02931d724c7b7ffffe17f72c85bc90190399085 - shared/managed-agents-tools.md: SHA-256 9a9fd54ebb5e81eb0c6b7505f1cadc420fed5696cdab3e0080dd00ccf2e37b6f - shared/managed-agents-webhooks.md: SHA-256 38746f77a3ee971f9b1f7d47cec9e753131bf271fd795aed5c3da1764c9a6ac1 - shared/model-migration.md: SHA-256 b4fe96362b865a28726ab877c966624d5bfc2f86ff9b34940a34e3f8b6fc034e - shared/models.md: SHA-256 c1af1f093d46fd58e50ef3e8e28d339650f333059bfc49a63d01d0175675161b - shared/platform-availability.md: SHA-256 607997bc16de403970cb32c1dc1fa63e68f7f02328728be59397e7cefb2efed4 - shared/preserved-thinking-migration.md: SHA-256 29921d69fa8dc7e450e7bb793c9f59ae755262ccc933a52eb7116c207defb34c - shared/preserved-thinking-migration/causes.md: SHA-256 c2dbadb0439806863723e5ef7279d626d731d8179fd39293ff26d226b94d3694 - shared/preserved-thinking-migration/drop_block_probe.py: SHA-256 f23abb75bc079d835f57da7598ac5a1bee4318214ab803185feb94d8db8c4b88 - shared/preserved-thinking-migration/prefix_diff.py: SHA-256 734934b81b5f26e9ed2f8f51f44fed6554233de4f37b22a8845ae4ffcb622ac0 - shared/prompt-audit.md: SHA-256 1a50ec950f9c11bb4ad1a5032a26bb20cd52c9d358adf6add2cb1ceb1aaf6cc9 - shared/prompt-caching.md: SHA-256 5800f883ff1babe24105e7e0e3fc7e9b03dfcce25bb2ccded4bb40d8c81d8464 - shared/token-counting.md: SHA-256 8674e246c165decdc13106a4861a254ededb0421caf7fa4015ed0fb36e6c60dc - shared/tool-use-concepts.md: SHA-256 62293e7bb7763f7ae487d3b98ead6b733689b2fd7089784ad5859dd869a153f5 - typescript/claude-api/README.md: SHA-256 b2a01ccbaa0d4406e400bcd8efc033759af5d646a97aab744aee55bf07d6ab16 - typescript/claude-api/batches.md: SHA-256 acc33bb981acd48b6ee9a6b2c2afeef9487c42ef2b117225da3949dc61efb464 - typescript/claude-api/files-api.md: SHA-256 df8e4e433a4ded6d4c3eb40de9bc0782cf77aaa8e3e331a539aec1cff55d1316 - typescript/claude-api/streaming.md: SHA-256 09db982effb0b7d14e29eddd5425b00040debfedccf3dd0db9af73f7d2da3e79 - typescript/claude-api/tool-use.md: SHA-256 d0c3e5da146c6da8a85ced14f964dc33de4c0153aa2748dee35c319752e8b725 - typescript/managed-agents/README.md: SHA-256 5daf6b0c6a443e1ae720b0a7343e061ebdbaeb751dd674f37f9abf9d12efc66f If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, and tell me in plain words: what the skill does when used, what it connects to, whether it reads or sends any personal information or credentials, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The skill folder holds 96 files: the main SKILL.md, LICENSE.txt (Apache-2.0), per-language reference folders (csharp, curl, go, java, php, python, ruby, typescript) and a shared/ folder of topical guides. Four companions are code files (two Python probe/diff scripts and two Node .mjs report scaffolds inside the eval and preserved-thinking guides); they are reference material, copied as-is, and nothing in the folder is run during installation. - Nothing is sent anywhere during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. No Anthropic API call is made and no credential is read, asked for or stored during installation. - At use time, the code the agent writes with this reference calls the Anthropic API (or the cloud provider I chose, such as Bedrock, Vertex AI or Foundry) with my own authentication and is billed per token; what is sent then is the prompts and data my own application puts into its API requests. The skill may also fetch up-to-date official documentation (the live sources listed in shared/live-sources.md) when a binding is not in the bundled files. - The skill's own authentication rule: check the existing credential chain first (ant auth status, SDK resolution, login profile); ask me for a key only if no credential source is active. Admin features require a separate admin credential, and money-spending workflows (evals, replays, hill-climbing) require my budget approval before they run. - Do not run any script from the folder, do not call the API, do not check or set any credential, and do not use the skill in any way during this install: authentication, API calls and use are my decisions, later, when I actually ask for them. 3. When I say "yes, install", install it as a skill: copy the whole claude-api folder into the agent's skills directory, in a folder named exactly "claude-api", keeping its 96 files in their original relative layout (SKILL.md at its root). 4. Verify with no network calls: frontmatter valid (name: claude-api), the 96 files in place with matching fingerprints. Do not run anything from the folder and do not use the skill to verify. 5. Report what was installed, where, and what I still need to do myself (for example: that using what this skill builds needs my own active Anthropic authentication and is billed per token, that I should use my existing login profile or key instead of pasting a new one, and that money-spending workflows need my budget approval first; using the skill is a separate step that only starts when I ask for it, and the use-time API calls disclosed above happen only then). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.