codeflicker_skill

Overview

This skill is the CodeFlicker CLI programming assistant for fast, command-line code generation and editing. It provides interactive and quiet modes, model selection, and configuration for automated workflows. Use it to generate code, continue sessions, run tasks in a specified working directory, and configure execution behavior.

How this skill works

CodeFlicker runs as a global npm CLI that communicates with AI models to generate, modify, and execute code tasks. You log in once with SSO, set global configuration (models, approval mode), then invoke commands in interactive or quiet modes. The tool supports session continuation, model selection, and an "yolo" approval mode that auto-executes changes.

When to use it

• You want a quick CLI assistant to write or refactor code from the terminal.
• You need to continue or resume an AI-driven coding session non-interactively.
• You must run codegen tasks inside a specific project directory or scriptable pipeline.
• You want to switch models for lightweight editing or planning/vision tasks.
• You need an internal tool that uses SSO-based authentication for secure access.

Best practices

• Always run flickcli /login before first use to ensure authenticated access.
• Test changes locally before enabling approvalMode yolo; yolo will auto-execute edits.
• Use model flags (-m) to choose lightweight models for quick edits and planning models for complex tasks.
• Set --cwd to the project root to avoid unintended edits outside the target repository.
• Keep sessions explicit: use -q -c to continue prior work and -q -r <id> to resume specific sessions.

Example use cases

• Generate a project scaffold: flickcli "create a Flask app with tests" in the project directory.
• Add unit tests to an existing module non-interactively: flickcli -q -c "add unit tests".
• Switch to a lightweight model for quick changes: flickcli -m claude-haiku-4.5 "simplify function".
• Auto-execute routine maintenance (use with caution): set approvalMode to yolo and run repeatable refactors.
• Resume a prior code session by ID to pick up work: flickcli -q -r <id> "continue".