Skip to main content

Build with LLMs

Use LLMs in your Silent Witness integration workflow.

You can use large language models (LLMs) to assist in building Silent Witness integrations. We provide a set of tools and best practices if you use LLMs during development.

Use agent skills

Agent skills give LLMs task-specific instructions. Silent Witness offers two skills:

  • REST API skill: Complete reference for building integrations using HTTP/curl — covers endpoints, field names, request/response formats, analysis types, and async polling.
  • SDK skill: Complete reference for building integrations using the Go or TypeScript SDK — covers method signatures, StartAnalysis helper, field naming conventions, and error handling.

Adding skills to your platform

Install the skills:

mkdir -p .claude/skills/sw-rest-api .claude/skills/sw-sdk

curl -o .claude/skills/sw-rest-api/SKILL.md \
https://raw.githubusercontent.com/silentwitnessai/agent-skills/main/skills/sw-rest-api/SKILL.md

curl -o .claude/skills/sw-sdk/SKILL.md \
https://raw.githubusercontent.com/silentwitnessai/agent-skills/main/skills/sw-sdk/SKILL.md

Use the skills:

Invoke /sw-rest-api or /sw-sdk followed by what you want to build:

/sw-rest-api Write a Python script that creates a delta_v_only case,
uploads damage photos, and downloads the report as PDF
/sw-sdk Write a Go program that uses StartAnalysis to run a full
accident_injury analysis with occupant data

Skills are automatically detected from .claude/skills/ — no additional configuration needed.

See also