Publisher: Cloudflare plugin in openai/plugins · Editorial profile by AI Tools Ratings · Review method
What task is this skill designed for?
Use the source to connect an agent class, state transitions, client routing and background work. Separate deterministic state updates from model responses so retries do not silently repeat a business action.
What input and output should you expect?
Input: A small TypeScript project, an explicit state model, expected client events and a list of actions requiring approval. Use synthetic records in the first pass.
Output: A reviewable implementation plan and code changes for state, bindings, routing and client interaction, plus a checklist for reconnects and duplicate events.
What does it need to run?
A compatible coding agent, Node tooling, the agents package and Workers project configuration. Remote deployment needs a Cloudflare account; model integrations need their own access.
Published in the OpenAI Codex plugins repository. Compatibility evidence is the plugin package and its source instructions; this is not a claim of identical behavior in other agent clients.
How should you approach installation?
- Open the pinned source and review the complete skill folder, including referenced helpers and license, before enabling it.
- Use the corresponding plugin in a compatible Codex environment; preserve the skill folder structure rather than copying only SKILL.md. Confirm current plugin availability in your client.
- Start with a disposable project and the trial prompt below. Grant only the stated access, inspect the proposed diff, and require separate approval for remote writes or publishing.
Installation guide: discovery, dependencies and runtime checks
What files, accounts or data can it access?
Local code and configuration can be read or changed. Enabled cloud tools can create resources and send prompts or state to Cloudflare and a chosen model provider. Keep production credentials outside the trial.
What does the skill cost?
The instruction files have no metered execution fee. Coding-agent usage, Workers resources, storage and any model provider are separate costs; estimate retries and scheduled executions.
What are its limits and alternatives?
The skill is guidance for an SDK, not a hosted autonomous worker. Examples require current API checks. Persistent state does not by itself provide authorization or exactly-once external actions.
Use the broader Cloudflare platform skill for product selection, or Wrangler guidance for deployment configuration. Neither replaces application-level state design.
What could you ask in a small trial?
Try this with a small sample:
Design a local counter agent with two clients. Reject negative values, define reconnect behavior and describe how duplicate calls are handled. Produce the diff and tests to run; do not deploy or connect paid models.
How would you verify the result?
- A rejected update leaves persisted state unchanged.
- Two clients converge after disconnect and reconnect.
- Repeated action requests cannot duplicate an external operation.
- All public methods have an explicit access decision.
Use a baseline and record acceptance evidence before treating a successful task as a reliable workflow.
Which source and license were checked?
Read the reviewed SKILL.md at its pinned commit · Complete reviewed folder
Repository and current publisher guidance
Checked:
Apache-2.0 License evidence
Version details and review scope
Source: openai/plugins/plugins/cloudflare/skills/agents-sdk
Reviewed commit: 0722921d5542fc593105c27bd52630babd8b8c2a
SKILL.md SHA-256: 007d6ccf85808a6c779f5b058b1be970d77785000071cb27a55a5f580fbc845b
Pinned source and license review. This profile covers the source workflow, stated dependencies and license evidence. See our editorial methodology for how we distinguish source reviews and hands-on evaluations.