A collection of documentation and guides designed for AI agents working with LaunchDarkly.
These guides provide structured instructions to help AI agents effectively interact with LaunchDarkly through its CLI, APIs, and data payloads.
- launchdarkly-cli.md - Reference for AI agents using the LaunchDarkly CLI (
ldcli) to interact with LaunchDarkly resources - launchdarkly-docs.md - Instructions for searching and accessing LaunchDarkly documentation programmatically
- launchdarkly-fallback-audit.md - Instructions for auditing LaunchDarkly feature flag fallback values and providing recommendations based on the current state of targeting rules
- launchdarkly-flag-segment-rules.md - Comprehensive guide for extracting information from LaunchDarkly initialization payloads using
jq - launchdarkly-implementation-audit.md - Audit LaunchDarkly implement against best practices using the SDK Preflight Checklist
These prompt files are designed to be provided to AI coding agents (like Cursor, GitHub Copilot, or Claude) when working with LaunchDarkly. Each file contains:
- Structured instructions and workflows
- Command examples and templates
- Best practices for the specific task
To use a prompt:
- Copy the relevant prompt file content
- Provide it to your AI agent as context
- Follow the agent's guidance for the specific LaunchDarkly task
Install prompts as Cursor commands for quick access:
# Install all prompts as Cursor commands
just install-all-commands
# Or install specific prompts
just install-fallback-audit
just install-command launchdarkly-cli.mdAfter installation, use prompts in Cursor with /launchdarkly-fallback-audit, /launchdarkly-cli, etc.
Build and install as Claude Skills for use with Claude Code:
# Build all skills in ./skills/ directory
just build-all-claude-skills
# Install all skills to ~/.claude/skills/
just install-all-claude-skills
# Or install specific skills
just install-claude-skill launchdarkly-cliClaude will automatically discover and use installed skills based on your conversation context.
Available recipes:
add-to-toc FILE DESC # Add a single file with description to TOC
build-all-claude-skills # Build all agent prompts as Claude Skills into ./skills/
build-claude-skill FILE # Build a Claude Skill from a prompt file into ./skills/
build-claude-skills-project # Build all agent prompts as project Claude Skills (.claude/skills/)
clean-claude-skills # Remove installed Claude Skills from ~/.claude/skills/
clean-claude-skills-project # Remove project Claude Skills (.claude/skills/)
clean-skills-build # Clean the local ./skills/ build directory
install-all-claude-skills # Install all Claude Skills from ./skills/ to ~/.claude/skills/
install-all-commands # Install all agent prompt files as Cursor commands
install-claude-skill SKILL # Install a Claude Skill from ./skills/ to ~/.claude/skills/
install-command FILE # Install a specific prompt file as a Cursor command
install-fallback-audit # Install launchdarkly-fallback-audit as a Cursor command
rebuild-toc # Rebuild entire TOC from scratch (preserves existing descriptions, adds placeholders for new files)
uninstall-command FILE # Uninstall a Cursor command
update-readme-commands # Update the just commands list in README.md
update-toc # 4. Cursor agent will then use 'just add-to-toc FILENAME "Description"' to add them- jq - Required for initialization payload parsing (install via
brew install jqon macOS) - LaunchDarkly CLI (
ldcli) - Required for CLI operations (see installation guide) - just - Optional, for automation recipes (install via
brew install juston macOS)
Apache 2.0 - see LICENSE file for details.
This repository is maintained by LaunchDarkly Labs. While we try to keep it up to date, it is not officially supported by LaunchDarkly. For officially supported SDKs and tools, visit https://launchdarkly.com