Runs Ralph orchestrator with the code-assist configuration to implement code tasks autonomously. Spawns Ralph as a subprocess that handles the full TDD workflow (Explore, Plan, Code, Commit) for the given code task file.
Installation
Details
Usage
After installing, this skill will be available to your AI coding assistant.
Verify installation:
skills listSkill Instructions
name: ralph-code-assist description: Runs Ralph orchestrator with the code-assist configuration to implement code tasks autonomously. Spawns Ralph as a subprocess that handles the full TDD workflow (Explore, Plan, Code, Commit) for the given code task file. type: anthropic-skill version: "1.0"
Ralph Code Assist
Overview
This skill launches the Ralph orchestrator with the ralph.code-assist.yml configuration to implement code tasks autonomously. Ralph handles the full TDD workflow (Explore → Plan → Code → Commit) using its event loop and hat system.
Use this when you have a .code-task.md file and want Ralph to implement it end-to-end without manual intervention.
Parameters
- code_task (required): Path to a
.code-task.mdfile containing the task specification - config (optional, default: "ralph.code-assist.yml"): Ralph configuration file to use
- verbose (optional, default: false): Enable verbose output from Ralph
Constraints for parameter acquisition:
- You MUST ask for all required parameters upfront in a single prompt
- You MUST validate that the code_task file exists before proceeding
- You MUST verify the code task file has the
.code-task.mdextension or is a valid task file
Steps
1. Validate Inputs
Verify the code task file exists and is properly formatted.
Constraints:
- You MUST check that the code_task file path exists
- You MUST read the code task file to verify it contains valid task content
- You MUST notify the user if the file is missing or malformed
- You SHOULD display a summary of the task (name, description, complexity) before proceeding
2. Display Task Summary
Show the user what will be implemented before starting.
Constraints:
- You MUST display the task name, description, and complexity from the code task file
- You MUST list the acceptance criteria that will be implemented
- You MUST confirm the current working directory where changes will be made
3. Execute Ralph
Run the Ralph orchestrator with the code-assist configuration.
Constraints:
- You MUST run the command:
cargo run --bin ralph -- run -c {config} -p "Implement the task at: {code_task}" - You MUST capture and display Ralph's output in real-time or stream it to the user
- You MUST monitor for the completion promise "LOOP_COMPLETE" or error conditions
- You MUST handle timeouts gracefully (default max runtime is 8 hours)
- You SHOULD provide periodic status updates if Ralph runs for an extended time
4. Report Results
Summarize the execution results for the user.
Constraints:
- You MUST report whether Ralph completed successfully or encountered errors
- You MUST display the termination reason (completion promise, timeout, error)
- You MUST list any files created or modified during execution
- You SHOULD provide a summary of what was implemented if successful
- You SHOULD suggest next steps (review changes, run tests, commit, etc.)
Example Usage
/ralph-code-assist tools/task-01-replay-backend.code-task.md
Or with a different config:
/ralph-code-assist code_task=tools/task-01-replay-backend.code-task.md config=ralph.smoke.yml
Expected Output
When successful, Ralph will:
- Read the code task specification
- Follow the code-assist SOP (TDD workflow)
- Implement tests first, then implementation code
- Run tests to verify
- Commit changes with conventional commit message
- Emit "LOOP_COMPLETE" when done
Troubleshooting
Ralph Build Fails
If cargo run --bin ralph fails:
- You SHOULD run
cargo build --bin ralphfirst to see build errors - You SHOULD check that all dependencies are available
- You SHOULD verify the ralph.code-assist.yml config file exists
Task Not Completing
If Ralph runs but doesn't complete:
- You SHOULD check the scratchpad at
.agent/scratchpad.mdfor progress - You SHOULD verify the task requirements are clear and achievable
- You SHOULD consider if the task needs to be broken into smaller pieces
Agent Artifacts
Ralph creates working files during execution:
.agent/scratchpad.md- Ralph's working memory between iterations- These are normal and can be cleaned up after successful runs
- You SHOULD add
.agent/to .gitignore if not already present
Related
ralph.code-assist.yml- The Ralph configuration for code assist mode.sops/code-assist.sop.md- The SOP that Ralph follows (if exists)/code-assistskill - Direct Claude code assist (not via Ralph)
More by mikeyobrien
View allEvalKit is a conversational evaluation framework for AI agents that guides you through creating robust evaluations using the Strands Evals SDK. Through natural conversation, you can plan evaluations, generate test data, execute evaluations, and analyze results.
Use when testing Ralph's hat collection presets, validating preset configurations, or auditing the preset library for bugs and UX issues.
Use when bumping ralph-orchestrator version for a new release, after fixes are committed and ready to publish
Use when creating new Ralph hat collection presets, designing multi-agent workflows, or adding hats to existing presets
