codex-harnesses

Route an independent review-only agent to a stronger model tier or reasoning effort at key decision points, then continue the task in the main agent. Use when the user asks for an adviser, advisor, second opinion, stronger review, or Claude Code /advisor-like behavior; before committing to a consequential approach on a multi-step task; when work is stuck or changing direction; and before declaring substantial work complete.

Monitor or resume GitHub pull requests, fix CI failures and actionable CodeRabbit, Codex, bot, or human review feedback, push focused updates, and stop only when the PR is merge-ready or has a concrete external blocker.

Convert review feedback, CI failures, bug fixes, or task retrospectives into a durable project feedback ledger, then make future implementation work consult that ledger before planning.

Run an end-to-end implementation delivery workflow from user request to merge-ready pull request. Use when the user asks Codex to implement a change and expects planning, goal tracking, UI design where relevant, code edits, tests, self-review, regular ready-for-review PR creation, CI/review follow-up, and bringing the PR as close to mergeable as possible.

Analyze recent Codex work and package repeated patterns as the smallest useful skill, custom subagent, hook, or automation. Use when asked to mine sessions, memories, or recent repeated work for reusable Codex assets.

Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.

Remove common AI-generated boilerplate and readability noise from changed code without changing behavior.

Diagnose and repair failing CI checks on the current branch or pull request with focused, reproducible fixes.

Review a diff for concrete security risks across injection, authn/authz, secrets, supply chain, infrastructure, and business logic.

Remove avoidable complexity from changed code while preserving behavior exactly.

Drive implementation through red-green-refactor with isolated test design, minimal implementation, and behavior-preserving cleanup.

Read, summarize, inspect, or extract content from a public URL through Jina Reader. Use when the user asks to use Jina Reader, read a note.com/blog/article/X/Twitter/GitHub/page URL, recover Markdown from a URL that normal browsing cannot read cleanly, or make a URL readable in ChatGPT-like agent sessions with tool access.

Run a continuous product improvement loop: evaluate a product or codebase end to end, identify product-quality improvement opportunities, present prioritized recommendations for user approval, then implement only approved improvements through the plan, goal, UI design, testing, review, PR, and merge-readiness workflow. Use when the user asks Codex to assess an app/product, find what can be improved, run a product evaluation, or iteratively improve a product with approval gates.

Direct high-quality frontend UI work by using Codex Image Gen for visual mockups, then implementing the selected direction in code with browser-based visual QA. Use when the user asks for a better-looking web UI, UI mockup, app screen, dashboard, landing page, visual redesign, or wants Image Gen/GPT Image 2 to guide frontend implementation. Avoid for backend-only work or tiny style tweaks that do not need visual direction.

Fetch public web pages with Jina AI Reader API and convert them into LLM-friendly Markdown. Use when an agent needs to read, summarize, quote briefly, extract facts from, or inspect a public URL that normal browsing, scraping, search snippets, or direct fetch tools cannot read cleanly, including JavaScript-heavy pages, reader-hostile pages, X/Twitter posts and threads, articles, GitHub pages, and PDFs. Trigger when the user asks to use Jina, use r.jina.ai, read an otherwise unreadable site, recover readable text from a URL, read an X/Twitter post by URL, or turn a webpage into Markdown.

Methodology for iteratively improving agent-facing instructions (skills / slash commands / CLAUDE.md / code-gen prompts) by having a bias-free executor run them and evaluating two-sidedly (executor self-report + instruction-side metrics) until improvements plateau. Use after creating or revising a prompt or skill.

Manage /goal usage for concrete Codex work, including objective writing, success criteria, done criteria, long-running implementation, verification, PR creation, resumption, and completion decisions.

Review code changes with a bug-first, risk-focused workflow before merge or release.

Fix bugs with reproduction, minimal changes, regression coverage, and targeted verification.

Update documentation after code or behavior changes while keeping docs concise and relevant.

Implement a feature safely with planning, focused edits, tests, docs, and verification.

Refactor code without behavior changes using small mechanical steps and verification.

Check whether a change is ready to release by reviewing diff, tests, docs, risks, and migration notes.