Runs Codex CLI as background sub-agents for image generation, code analysis, and parallel review.
npx skills add https://github.com/yan-labs/yan-skills --skill codex-yan-labs
Please help me install this Agent Skill. Skill: codex Source: https://github.com/yan-labs/yan-skills/tree/main/codex Command: npx skills add https://github.com/yan-labs/yan-skills --skill codex-yan-labs
codex exec with run_in_background, stdin-piped prompt files, sandbox selection, and 2>/dev/null stderr suppression so thinking tokens never flood the main context. - Parallel Agent Teams: Split work into independent slices (security review, perf review, migration drafts) across multiple Codex workers with collision-avoidance rules for workspace-write sandboxes. - Use Case: Ask for a set of 6 character illustrations for a landing page; the Skill writes a prompt file with exact filenames, palette, and sizing rules, launches Codex in the background, then verifies every image before wiring it into the page. ## Quick Start Ask the agent to use Codex to generate a set of images or to review a repository in the background, and it will launch codex exec with the right sandbox and effort level automatically.High-intent search queries and answers about installing and using this skill.
Launch `codex exec` with `run_in_background: true`, pipe the prompt via stdin from a temp file, and append `2>/dev/null` to suppress thinking tokens. Read results later with BashOutput when the completion notification arrives instead of polling.
Yes. Codex has a built-in OpenAI image generation tool as an agent capability, not a CLI subcommand. Describe the images in the prompt with exact filenames, style, and palette, and run with `--sandbox danger-full-access` since generation needs network access.
Split the task into independent slices, write a prompt file per slice, and launch separate background Bash calls in the same message with distinct tags. Never run two `workspace-write` workers on overlapping paths; use disjoint directories or separate git worktrees.
No. By default do not pass `-m` or `--model`; Codex reads its default model from `~/.codex/config.toml`. Only add an explicit model flag when the user names a specific model in the request.
Codex streams thinking tokens to stderr, which bloats the main context window. Always append `2>/dev/null` to suppress them unless you are deliberately debugging Codex's reasoning process.
Open every image with Read before shipping it, then run three checks: a 120px thumbnail contact sheet for composition, an alpha bounding-box ratio for subject size, and an independent blind review for style compliance. Compress PNGs to JPEG before committing.
PsychQuantQuery OpenAI Codex CLI source code and documentation for configuration and usage details.
sanskarbasnetRun Codex CLI reviews, adversarial tests, and consult sessions with continuity.
fanciveCoordinate cross-model code reviews between Claude and Codex with structured P1/P2/P3 reports.
bigbulgogiburgerAutomate raster image creation and editing via Codex's image_gen tool.