codeconductor

SKILL.md content

Unable to load SKILL.md content from source.

How this skill works

• Triggered by task category 'code'
• Uses llm tool
• Parses the task input to identify language, framework, and key requirements
• Uses LLM to create a detailed plan including file structure, modules, and implementation strategy
• Generates initial code skeleton based on the plan
• Fills in the code skeleton with functional logic
• Generates unit tests for the code
• LLM reviews the code for errors, best practices, and improvements
• If issues found, iterates on previous phases
• Formats final code with README, comments, and execution instructions
• Maintains internal state tracking current phase and previous outputs
• Each phase uses context-aware prompts referencing prior outputs
• Retries failed LLM calls up to 3 times
• Validates generated code syntax before final output

When to use it

• When tasked with solving a programming problem requiring code generation, execution, and debugging in Python
• When needing to iteratively develop and test code in a sandboxed REPL environment to achieve a computational goal

Best practices

• Set either OPENAI_API_KEY or ANTHROPIC_API_KEY environment variable
• Secure API keys using environment variables, not hardcoded
• Be specific in prompts including language, frameworks, and requirements
• Provide context by referencing existing files or previous conversations
• Build incrementally by iterating on smaller tasks
• Always review code changes before approving execution
• Use natural language as if talking to a senior engineer
• Approve terminal commands only after review
• Test generated code thoroughly

Example use cases

• Prototyping applications: Agents prototype applications by generating, editing, executing, and testing code in isolated environments.
• Data analysis: Agents perform data analysis tasks using code generation, dependency installation, and execution for Python scripts.
• Algorithm implementation: Agents implement and test algorithms by creating code files, running them, and iterating based on output.
• Complex programming tasks: Agents handle complex programming tasks through multi-file project management, editing, debugging, and sandboxed execution.

FAQs