Anthropic

Anthropic Claude API integration — chat completions, streaming, vision, tool use, and batch processing via the Anthropic Messages API. Generate text with Cla...

Security Scan

The skill claims full Anthropic Claude features (chat, streaming, vision, tool calling), but the included script appears to be a lightweight/stub implementation that does not actually implement key capabilities and also reads a fallback .env file — the overall package is inconsistent with its claims.

Purpose & Capability

Name/description promise: complete Anthropic Messages API support including streaming, image analysis, tool use, and batch processing. What is present: a single stdlib Python CLI that only issues simple HTTP requests (GET/POST) to a set of constructed endpoints. Several claimed capabilities (streaming, image upload/analysis, tool calling with structured input, real message sends) are not properly implemented: e.g., image handling and streaming would normally require multipart uploads or chunked reads, but the code uses urllib.request.urlopen(read()) with no file upload or streaming logic. Many CLI commands that one would expect to POST data instead call GET endpoints. This mismatch suggests the code is a stub or incomplete and does not deliver the full features the SKILL.md advertises.

Instruction Scope

SKILL.md instructs the agent to run commands that imply uploading images, streaming responses, and tool-enabled chats. The runtime instructions do not disclose that the script will fall back to reading a .env file in a workspace path if ANTHROPIC_API_KEY is not set. The script reads WORKSPACE (if set) or ~/.openclaw/workspace/.env to find the key, which is broader local-file access than the SKILL.md explicitly highlights. Also the instructions assume the CLI will implement features it apparently does not (streaming, image analysis), giving an overbroad impression of capabilities.

Install Mechanism

No install spec; instruction-only plus a single Python stdlib script. No external downloads or package installs. This is low-risk from an installation/execution-files perspective.

Credentials

Declared requirement: ANTHROPIC_API_KEY (primary credential) — appropriate for an Anthropic integration. Implementation detail: the script will also consult a WORKSPACE env var (if present) and attempt to read a .env file in that workspace (~/.openclaw/workspace/.env by default) to extract ANTHROPIC_API_KEY. While the script only attempts to parse a line starting with ANTHROPIC_API_KEY= (so it isn't indiscriminately reading other secrets), the fallback file access is broader than the simple single-env-var requirement and may surprise users who store other credentials in that .env file.

Persistence & Privilege

Skill is not always-included and uses default agent invocation settings. It does not attempt to modify other skills or agent-wide configuration. It does read a local file as a fallback for the API key, but it does not persist credentials itself.

What to consider before installing

This skill is not clearly dishonest, but it is inconsistent. Before installing: 1) Inspect and test the included script in an isolated environment — it looks like a stub and may not actually perform streaming, image uploads, or tool-calling as advertised. 2) If you rely on the advertised features (streaming, vision, tool use), prefer the official SDK or a more complete implementation; do not assume these work. 3) Be aware the script will try to read ANTHROPIC_API_KEY from a fallback ~/.openclaw/workspace/.env (or a WORKSPACE path you set). If you store other secrets in that .env, move them or supply the API key via environment variable to avoid accidental exposure. 4) If you need higher assurance, ask the author for an explanation of the GET/POST semantics and a demonstration of image upload and streaming, or request source from the claimed GitHub link and validate network calls. If you cannot validate those, treat this skill as incomplete and avoid using it with production secrets.

Anthropic Claude API integration — chat completions, streaming, vision, tool use, and batch processing via the Anthropic Messages API.

Features

• Messages API — Claude Opus, Sonnet, Haiku completions
• Streaming — real-time token streaming responses
• Vision — image analysis and understanding
• Tool use — function calling with structured output
• System prompts — custom system instructions
• Multi-turn conversations — context management
• Batch API — bulk message processing
• Token counting — estimate usage before sending
• Extended thinking — deep reasoning mode
• Model listing — available models and capabilities

Requirements

Variable	Required	Description
ANTHROPIC_API_KEY	✅	API key/token for Anthropic

Quick Start

# Send a message to Claude
python3 {baseDir}/scripts/anthropic.py chat "What is the meaning of life?" --model claude-sonnet-4-20250514

# Chat with system prompt
python3 {baseDir}/scripts/anthropic.py chat-system --system "You are a financial analyst" "Analyze AAPL stock"

# Analyze an image
python3 {baseDir}/scripts/anthropic.py chat-image --image photo.jpg 'What do you see in this image?'

# Stream a response
python3 {baseDir}/scripts/anthropic.py stream "Write a short story about a robot" --model claude-sonnet-4-20250514

Commands

chat

Send a message to Claude.

python3 {baseDir}/scripts/anthropic.py chat "What is the meaning of life?" --model claude-sonnet-4-20250514

chat-system

Chat with system prompt.

python3 {baseDir}/scripts/anthropic.py chat-system --system "You are a financial analyst" "Analyze AAPL stock"

chat-image

Analyze an image.

python3 {baseDir}/scripts/anthropic.py chat-image --image photo.jpg 'What do you see in this image?'

stream

Stream a response.

python3 {baseDir}/scripts/anthropic.py stream "Write a short story about a robot" --model claude-sonnet-4-20250514

batch-create

Create a batch request.

python3 {baseDir}/scripts/anthropic.py batch-create requests.jsonl

batch-list

List batch jobs.

python3 {baseDir}/scripts/anthropic.py batch-list

batch-get

Get batch status.

python3 {baseDir}/scripts/anthropic.py batch-get batch_abc123

batch-results

Get batch results.

python3 {baseDir}/scripts/anthropic.py batch-results batch_abc123

count-tokens

Count tokens in a message.

python3 {baseDir}/scripts/anthropic.py count-tokens "How many tokens is this message?"

models

List available models.

python3 {baseDir}/scripts/anthropic.py models

tools

Chat with tool use.

python3 {baseDir}/scripts/anthropic.py tools --tools '[{"name":"get_weather","description":"Get weather","input_schema":{"type":"object","properties":{"location":{"type":"string"}}}}]' "What is the weather in NYC?"

thinking

Extended thinking mode.

python3 {baseDir}/scripts/anthropic.py thinking "Solve this math problem step by step: what is 123 * 456?" --budget 10000

Output Format

All commands output JSON by default. Add --human for readable formatted output.

# JSON (default, for programmatic use)
python3 {baseDir}/scripts/anthropic.py chat --limit 5

# Human-readable
python3 {baseDir}/scripts/anthropic.py chat --limit 5 --human

Script Reference

Script	Description
{baseDir}/scripts/anthropic.py	Main CLI — all Anthropic operations

Data Policy

This skill never stores data locally. All requests go directly to the Anthropic API and results are returned to stdout. Your data stays on Anthropic servers.

Credits

---

Built by M. Abidi | agxntsix.ai YouTube | GitHub Part of the AgxntSix Skill Suite for OpenClaw agents.

📅 Need help setting up OpenClaw for your business? Book a free consultation