- Home
- Skills
- Tdhopper
- Dotfiles2.0
- Sending To Codex
sending-to-codex_skill
- Shell
3
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
3 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 tdhopper/dotfiles2.0 --skill sending-to-codex- SKILL.md4.3 KB
Overview
This skill delegates coding tasks and technical questions to OpenAI's Codex CLI from within Claude Code. It supports fire-and-forget edits (fixes, refactors, features) and non-destructive analysis or research queries. Use clear intent phrases like "ask codex" or "send to codex" to trigger delegation.
How this skill works
The skill selects the appropriate Codex mode based on intent: full-auto edits for file changes, ephemeral/read-only for analysis, or codex review for code reviews. It builds a codex exec or codex review command with flags for working directory, output file, sandboxing, and model overrides, then runs the command and collects results. After execution it summarizes changes or reads the output file to relay Codex's response.
When to use it
- Fix a bug, add a feature, or refactor code and you want Codex to make changes automatically
- Ask a research or design question about the codebase and you only want an analysis or explanation
- Request a code review of uncommitted changes or against a base branch
- Offload long-running or noisy editing tasks to a separate agent
- Get a second opinion on architecture, dependency issues, or potential vulnerabilities
Best practices
- Determine intent first: editing (use --full-auto) vs analysis (use --ephemeral or -s read-only)
- Provide a specific, actionable prompt with relevant file paths and constraints
- Set -C to the correct working directory so changes land in the intended worktree
- For long or complex instructions, pipe a multi-line prompt from stdin to preserve clarity
- After edits, inspect git diff in the target directory before accepting or summarizing results
Example use cases
- Fix a NameError in src/utils.py: run codex exec --full-auto -C /path "Fix the broken import causing NameError in parse_config()"
- Explain complex code: codex exec --ephemeral -o /tmp/out.md -C /repo "Explain how the converter DAG works and highlight circular dependencies"
- Review uncommitted changes for security issues: codex review --uncommitted -C /repo
- Refactor a module to use an existing pattern: codex exec --full-auto -C /repo "Refactor X to follow pattern Y; do not modify tests"
- Run a long transformation in background and check back later for git diff
FAQ
Use --ephemeral and -s read-only when running codex exec, or write output to a file with -o to capture analysis only.
Which flag makes Codex apply changes automatically?
Use --full-auto to enable sandboxed automatic execution so Codex can modify files without interactive approval.