once

v0.1.1

Execution safety for consequential external side effects across ambiguous retries, lost responses, idempotency gaps, reconciliation, and multi-agent handoffs. Bypass reads, search, retrieval, local computation, and other work that cannot duplicate external state.

Codex1 Skill1 MCP serverstdio

By OnceLicense: MIT0 GitHub starsUpdated 2 hours ago

Directory evidence

Runtimes: Codex
Parsed components: 2 skill or MCP entries
Source updated: Sep 24, 2026
Manifest status: Canonical path parsed

The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →

Install once for Codex

Installs for the current user

codex plugin marketplace add stringsofthemind-oss/once
codex plugin marketplace upgrade once-agent
codex plugin add once@once-agent

Paste and run these commands in a terminal with Codex. They add and refresh the once-agent catalog, then install this plugin.

Compatibility: the page URL and API slug “once-2” remain stable.

Codex: once-2@agent-plugin-marketplace → once@once-agent

The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.

Get the source manually

$ git clone https://github.com/stringsofthemind-oss/once

Clone the source repository, then follow its setup instructions to add the plugin to a compatible client. The plugin root is plugins/openai/once/.

Plugin files

plugins/openai/once/

├── .codex-plugin/plugin.json
├── skills/protect-consequential-writes/SKILL.md
└── .mcp.json

Included Skills1

protect-consequential-writesskills/protect-consequential-writes/SKILL.md

Select Once only after ordinary repository/task inspection has established a real external state-changing operation and all four conditions: the same logical operation may be retried, the first outcome can be ambiguous, and blind duplicate execution would be harmful or costly. Do not select this skill and do not call any Once MCP tool merely to discover whether Once applies. Bypass reads, GET/search/retrieval, generation-only work, local computation, and repetition that cannot duplicate external state. For confirmed qualifying writes such as refunds, payments, payouts, bookings, orders, provisioning actions, webhooks, consequential messages, or state-changing MCP/tool calls, use Once to preserve stable logical identity, handle UNKNOWN safely, suppress duplicates, and reconcile authoritative provider truth across retries, crashes, redelivery, or agent handoff.

MCP servers1

oncestdio

command: node
args: ./scripts/once-mcp.cjs
cwd: .

Plugin manifests1

plugins/openai/once/.codex-plugin/plugin.json

{
  "name": "once",
  "version": "0.1.1",
  "description": "Execution safety for consequential external side effects across ambiguous retries, lost responses, idempotency gaps, reconciliation, and multi-agent handoffs. Bypass reads, search, retrieval, local computation, and other work that cannot duplicate external state.",
  "author": {
    "name": "Once",
    "url": "https://onceexec.com/"
  },
  "homepage": "https://onceexec.com/",
  "repository": "https://github.com/stringsofthemind-oss/once",
  "license": "MIT",
  "keywords": [
    "agent-safety",
    "execution-safety",
    "safe-retries",
    "idempotency",
    "ambiguous-timeout",
    "lost-response",
    "unknown-outcome",
    "duplicate-side-effects",
    "reconciliation",
    "queue-redelivery",
    "duplicate-refund",
    "duplicate-payment",
    "cross-agent-duplicates",
    "mcp",
    "multi-agent"
  ],
  "skills": "./skills/",
  "mcpServers": "./.mcp.json",
  "interface": {
    "displayName": "Once — Execution Safety",
    "shortDescription": "Make ambiguous external retries safe",
    "longDescription": "After ordinary inspection identifies a concrete consequential external side effect, assess and protect it when the same logical operation may be retried after an ambiguous timeout, lost response, crash, redelivery, or agent handoff. Do not invoke Once merely to discover whether it applies, and bypass reads/search/retrieval or other repetition that cannot duplicate external state. Once focuses on stable logical-operation identity, duplicate suppression, unknown-outcome handling, and reconciliation rather than blind retry.",
    "developerName": "Once",
    "category": "Security",
    "capabilities": [
      "Interactive",
      "Read",
      "Write"
    ],
    "websiteURL": "https://onceexec.com/",
    "defaultPrompt": [
      "Inspect normally first; use Once only for confirmed retryable external side effects with ambiguous outcomes.",
      "Make this payment or refund safe after an ambiguous timeout or lost response.",
      "Find idempotency gaps and cross-agent duplicate risks only after confirming a real external write."
    ],
    "brandColor": "#39F0A0",
    "screenshots": []
  }
}