Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/logicrw/agpair/codexnpx skills add logicrw/agpair --skill codexgit clone --depth 1 https://github.com/logicrw/agpairWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00049 | $0.03912 |
| Opus 5 | $0.00024 | $0.01956 |
| Sonnet 5 | $0.00010 | $0.00782 |
| Haiku 4.5 | $0.00005 | $0.00391 |
Grade A, and why
agpair-codex scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 322 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGPair for Codex
IRON LAW: Non-trivial work requires an explicit routing budget. Use external AGPair lanes by default, but fan out only when each lane has a distinct role, scope, hypothesis, or verification value. Multiple grok-cli lanes are valid when they are not duplicate prompts.
Default: external AGPair executor first for non-trivial work. Codex remains the controller and verifier.
Actively outsource low-value, repetitive, time-consuming, or easily verifiable work through AGPair: repo scans, alternative reviews, focused test-fix attempts, multi-file mechanical edits, smoke checks, and implementation slices with clear acceptance criteria. For ordinary non-trivial work, start with a strong external lane: grok-cli and antigravity-cli are peer first choices. Use both when independent evidence, implementation contrast, or latency hedging is useful. It is acceptable to run several attempts from the same executor at once for parallel review, competing implementation approaches, or file-sliced work, but each attempt needs a distinct task id, role, prompt, or scope.
Use direct Codex edits for tiny local fixes, sensitive judgment-heavy work, or when AGPair is unavailable. If skipping external AGPair entirely for non-trivial work, state the skip reason before proceeding. Use Codex native subagents as narrow controller-side reviewers/helpers when they add clear value, and as fallback when external executors are unavailable, unsuitable, or not good enough.
Normal Task
For ordinary tasks, send a clear natural brief. AGPair normalizes useful
briefs and should not reject work merely because a section heading is missing.
Do not pass placeholders like <brief>, todo, or fix this.
External lanes do not share Codex's hidden conversation state. The brief must stand alone: include the goal, relevant paths, constraints, forbidden scope, expected output, and validation or evidence requirements. If the worker would need unstated chat context to succeed, either add that context to the brief or keep the work in Codex/native helper lanes.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 322 lines · 49 tokens per session scan A 256d05b76749
agpair-codex is a skill published in the GitHub repository logicrw/agpair (48 stars, last pushed 2mo ago), licensed MIT. It adds 49 tokens to every session and 3,912 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
data-visualization
Use for creating publication-quality charts and multi-panel analysis summaries. Triggers when tasks involve visualizing data, plotting results, creating charts, or producing visual reports from analysis output.
cuml-machine-learning
Use for GPU-accelerated machine learning on tabular data using NVIDIA cuML. Triggers when tasks involve classification, regression, clustering, dimensionality reduction, or model training on datasets.
blog-post
Writes and structures long-form blog posts, creates tutorial outlines, and optimizes content for SEO with cover image generation. Use when the user asks to write a blog post, article, how-to guide, tutorial, technical writeup, thought leadership piece, or long-form content.
social-media
Drafts engaging social media posts, writes hooks, suggests hashtags, creates thread structures, and generates companion images. Use when the user asks to write a LinkedIn post, tweet, Twitter/X thread, social media caption, social post, or repurpose content for social platforms.
remember
Review the current conversation and capture valuable knowledge — best practices, coding conventions, architecture decisions, workflows, and user feedback — into persistent memory (AGENTS.md) or reusable skills. Use when the user says: (1) remember this, (2) save what we learned, (3) update memory, (4) capture…
textual-screenshot
Capture a Textual terminal UI as an SVG using its headless test harness. Use when asked to make, attach, or preview a screenshot of deepagents-code/dcode or another Textual app, visually verify a TUI state, or render a modal, screen, or widget without a desktop or browser.