codeagent_skill

This skill orchestrates multi-backend code tasks with codeagent-wrapper, enabling parallel execution, file references, and structured outputs across Codex,

Go

2.1k

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill cexll/myclaude --skill codeagent

SKILL.md6.6 KB

Overview

This skill runs the codeagent-wrapper to execute multi-backend AI code tasks with configurable backends (Codex, Claude, Gemini), file references using @ syntax, and structured outputs. It orchestrates single and parallel workflows, supports session resume, and exposes environment controls for timeouts and concurrency. The goal is reliable, auditable automation for code analysis, generation, and refactoring.

How this skill works

You invoke the wrapper with a required --backend selection and provide a task or HEREDOC content; tasks may reference files with @file syntax. Parallel mode accepts a multi-task document where each task can override backend, workdir, and dependencies. Outputs are returned in either a summary (default) or full mode, and sessions can be resumed by passing a session ID to resume commands.

When to use it

Deep code analysis and large-scale refactoring across many files
Automated code generation that needs backend-specific strengths (analysis, docs, UI)
Running multiple dependent code tasks in parallel with per-task backend selection
Resuming long-running or multi-step agent sessions without losing context
Debugging complex workflows with full-output mode

Best practices

Start with Codex for heavy analysis, switch to Claude for documentation, and Gemini for UI tasks
Use HEREDOC syntax for long task prompts and file references (@path/to/file)
Limit concurrency in production by setting CODEAGENT_MAX_PARALLEL_WORKERS (recommended <= 8)
Avoid killing codeagent processes; check task logs and wait with a reasonable timeout instead
Enable Claude permission checks by default; set CODEAGENT_SKIP_PERMISSIONS only when trusted

Example use cases

Analyze a module call graph under @src/core and refactor inter-module dependencies using Codex
Generate comprehensive README and API docs for package files using Claude
Scaffold responsive UI components and layout prototypes with Gemini
Run a pipeline of analysis -> architecture design -> implementation with per-task backend selection
Resume an interrupted session to perform follow-up tasks against the same context

FAQ

Use the @file syntax (for example @src/main.go) inside your task content; the wrapper will include the referenced files in the agent context.

What output mode should I use for production?

Use the default summary mode for concise, context-efficient reports; enable --full-output only for debugging specific failures.