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 skills add e-t-y-b/etyb-skills --skill openaigit clone --depth 1 https://github.com/e-t-y-b/etyb-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/e-t-y-b/etyb-skills/openai)<a href="https://agentmods.dev/skills/e-t-y-b/etyb-skills/openai"><img src="https://agentmods.dev/badge/skills/e-t-y-b/etyb-skills/openai/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/e-t-y-b/etyb-skills/openai"><img src="https://agentmods.dev/badge/skills/e-t-y-b/etyb-skills/openai.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00263 | $0.04199 |
| Opus 5 | $0.00131 | $0.02099 |
| Sonnet 5 | $0.00053 | $0.00840 |
| Haiku 4.5 | $0.00026 | $0.00420 |
Grade A, and why
stack-openai 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 11d 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 — 132 lines — stays where its author put it; the contents beside it link to each section on GitHub.
OpenAI Stack — Team Briefing
This is a knowledge overlay, not a new specialist. The existing ETYB team does the work — backend-architect writes the backend code, devops-engineer wires the deploys, security-engineer enforces the boundary. This pack tells each role where the current OpenAI knowledge lives.
Where the full briefing lives
The full Stack briefing lives in this same folder. Per-product and per-role pages are siblings of this SKILL.md. Every page carries last_verified_on stamps and authoritative-source URLs in its frontmatter; see skills/etyb/core/knowledge-currency.md for the drift-check protocol that uses them.
- Stack briefing:
stacks/openai/index.md - Per-product pages:
stacks/openai/<product>.md— one per entry inproducts_coveredabove - Per-role views:
stacks/openai/<role>.md— one per role inapplies_to_rolesabove
When ETYB is installed locally these are read directly from disk. For third-party agents without the install, the same content is reachable as raw markdown at https://raw.githubusercontent.com/e-t-y-b/etyb-skills/main/stacks/openai/<page>.md.
When delegate_to_skills (frontmatter above) lists a first-party vendor MCP/skill that's installed in the user's environment, ETYB defers to it first. The in-repo Stack content is the curated fallback.
What changed in 2025-2026 that older training data misses
Critical context — an LLM with a 2024 cutoff will get these wrong:
- GPT-5 launched (2025) and replaced GPT-4 as the default recommendation. Tiers are GPT-5 Pro, GPT-5 Standard, GPT-5 Mini, GPT-5 Nano — plus thinking variants for the Pro/Standard tier. If a user types "use GPT-4 / gpt-4-turbo / gpt-4o" for a new feature, that is a legacy default; offer GPT-5 Standard or GPT-4.1 with a one-line rationale.
- The Responses API is the new unified surface (
/v1/responses). It replaces Assistants and is the only surface that supports built-in tools (web_search,file_search,code_interpreter,computer_use_preview) + remote MCP servers. Chat Completions is still supported long-term, but Responses is the right default for new agentic builds. - Assistants API is being deprecated. OpenAI announced the migration glide path in 2025; the sunset is scheduled in the first-half-2026 window. Do not greenfield on Assistants. Migrate threads → conversations, tools → built-in tools or function tools.
- OpenAI Codex (2025) is an AGENT PRODUCT, not the retired code model. The 2023
code-davinci-002"Codex" was retired in March 2023. The 2025 "OpenAI Codex" is a cloud + IDE coding agent (codex.openai.com+codexCLI) powered by GPT-5 family / o-series. - Computer Use Preview (consumer surface: Operator) lets the model drive a browser or desktop via screenshots + click/type tool calls. The API exposes it as the
computer_use_previewtool on Responses + thecomputer-use-previewmodel. The safety surface is large — sandboxing, allowlists, human-in-loop confirmation are required. - Prompt Caching is automatic on the OpenAI platform. No manual
cache_controlbreakpoints (this is distinct from Anthropic). Prompts ≥ 1,024 tokens that share a prefix with a recent prompt get cached; cached input is billed at 50% off. - The Agents SDK (Python + TypeScript) is the rebranded + hardened Swarm. It is the OpenAI-native answer to multi-agent orchestration: handoffs, guardrails, tracing, deterministic tool routing.
openai-agents-pythonis the import path. - Realtime API is GA and now ships Realtime Agents. Speech-to-speech with
gpt-realtime/gpt-4o-realtime, WebRTC for low-latency in-browser, WebSocket for server-side. - Structured Outputs with
strict: trueis the production default for any JSON. It enforces the JSON schema at decode time — no parsing failures, no malformed JSON. - Stored Completions + Eval Platform + Distillation Platform are three coupled console products. Turn on
store: trueon a completion → query it in the Eval platform → distill into a fine-tuned smaller model. - Project-scoped API keys (
sk-proj-…) replace user-scoped keys as the production pattern. Every production deployment should use a project key with a model allowlist, a per-project rate limit, and the project's own service account / audit log scope.
What ships with it
38 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents-platform.md 5.4 KB
- agents-sdk.md 7.8 KB
- ai-ml-engineer.md 14 KB
- assistants-api-legacy.md 7.4 KB
- audit-logs.md 5.4 KB
- backend-architect.md 14 KB
- batch-api.md 6.2 KB
- built-in-tools.md 8.3 KB
- chat-completions.md 8.4 KB
- codex-cli.md 4.4 KB
- computer-use.md 8.8 KB
- distillation-platform.md 5.8 KB
- embeddings.md 6.0 KB
- eval-platform.md 6.1 KB
- files-api.md 5.1 KB
- function-calling.md 8.3 KB
- gpt-4-1.md 5.6 KB
- gpt-5.md 8.8 KB
- image-generation.md 5.6 KB
- index.md 15 KB
- moderation-api.md 5.9 KB
- o-series-reasoning.md 7.9 KB
- openai-codex.md 5.5 KB
- openai-platform-console.md 5.6 KB
- organization-project-hierarchy.md 6.6 KB
- predicted-outputs.md 4.7 KB
- prompt-caching.md 6.4 KB
- realtime-agents.md 5.9 KB
- realtime-api.md 8.4 KB
- responses-api.md 9.5 KB
- security-engineer.md 20 KB
- stored-completions.md 6.0 KB
- structured-outputs.md 6.6 KB
- system-architect.md 18 KB
- tts.md 4.6 KB
- vision-fine-tuning.md 5.1 KB
- vision-input.md 5.9 KB
- whisper.md 5.3 KB
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.
- 11d ago First seen · 132 lines · 263 tokens per session scan A ccd51c183014
stack-openai is a skill published in the GitHub repository e-t-y-b/etyb-skills (10 stars, last pushed 2mo ago), licensed MIT. It adds 263 tokens to every session and 4,199 once invoked, about $0.0013 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 2026-08-31.
Other skills, from other repositories
portaljs-architect
Recommend a data-portal architecture (storage, compute, catalog, access, hosting, metadata) from stated needs, then hand off to the build skills. The advisory entry point. Use when starting a new data-portal project and the underlying architecture has not yet been decided.
map
Build and commit a Cortex function knowledge graph — maps structural dependencies and domain intent relationships across all AI functions in the project. Supports --reduce (default on) for transitive reduction of the dependency graph.
context-add
Guided wizard to capture or update organizational Snowflake/Cortex standards into .cocoplus/context/ .md. Presents a menu of 6 standard types, collects answers via multi-turn dialogue, and commits the file.
cocolean
CocoLean — minimum viable Cortex surface discipline. Pre-build decision ladder and intensity mode management. Invoke with $lean, $lean lite, $lean full, or $lean ultra.
test
Enter the Test phase of CocoBrew. Reads spec.md test requirements, generates test cases, executes SQL validation and quality checks, records results in test.md. Can be re-run without full rebuild. Requires Build phase completion.
map-diff
Analyze the impact of staged git changes against the committed Cortex function knowledge graph — shows which downstream functions are affected before you commit.