This plugin packages the deployed-network Iroha workflow for Codex around native Torii MCP.

It is intentionally optimized for live SORA/Torii networks such as Taira and future Nexus deployments, not contributor-local repo workflows.

• .codex-plugin/plugin.json — plugin metadata for Codex
• .mcp.json — built-in Taira MCP preset
• skills/iroha-live-network/ — guidance for safe live-network usage

The bundled MCP preset currently targets the primary public Taira endpoint:

• https://taira.sora.org/v1/mcp

If your deployment or operator gives you a different public Torii root for the environment under test, override it locally instead of editing the committed repo preset.

That endpoint must be enabled by the deployed validator config before the plugin can be used. If it returns 404, the node has not been redeployed with native Torii MCP yet.

For public write readiness, MCP discovery is not enough. A healthy rollout also needs a signed canary write to succeed; otherwise public reads can work while transactions still fail with route_unavailable.

Use the repo-local marketplace entry in .agents/plugins/marketplace.json and install the iroha plugin through Codex.

The plugin assumes the repo is the source of truth and keeps the current primary public Taira root committed. Alternate public roots should stay user-local.

This repo also ships standalone skills for the Codex Skills surface:

• skills/sora-taira-testnet/ for Taira testnet workflows
• skills/sora-minamoto-mainnet/ for Minamoto mainnet workflows

To install a skill from a GitHub checkout of this repo, use the built-in installer script from your local Codex environment and pass the desired skill path:

python3 "${CODEX_HOME:-$HOME/.codex}"/skills/.system/skill-installer/scripts/install-skill-from-github.py \
  --repo <owner>/<repo> \
  --path skills/sora-minamoto-mainnet

Restart Codex after installation so the skill appears in the Skills tab.

Public Taira nodes and custom Nexus/Torii networks are intentionally user-local rather than committed to the repo. Add them with a local MCP entry, for example:

codex mcp add iroha-custom --url https://<torii>/v1/mcp

Keep any network-specific auth headers, bearer tokens, or endpoint overrides in your local Codex config.

The deployed public profile this plugin expects is:

• torii.mcp.enabled = true
• torii.mcp.profile = "writer"
• torii.mcp.expose_operator_routes = false
• torii.mcp.allow_tool_prefixes = ["iroha."]

This makes Codex see the curated iroha.* aliases instead of the broader raw torii.* OpenAPI-derived surface.

Some iroha.* write-oriented tools accept JSON request bodies that include fields such as authority and private_key, or they submit pre-signed transaction envelopes.

Those values are runtime-only inputs:

• do not store them in repo config
• do not commit them into plugin manifests or docs
• do not write them to files unless the user explicitly asks for that
• prefer read-only queries until the user clearly asks to mutate live state

When live Taira writes fail with route_unavailable, treat that as an ingress or authoritative-peer deployment issue and rerun configs/soranexus/taira/check_mcp_rollout.sh --public-root https://taira.sora.org --write-config <runtime-only client.toml> or the exact public Torii root under test rather than debugging the plugin surface first. If the deploy still relies on hand-edited validator configs, rebuild them from configs/soranexus/taira/validator_roster.example.toml plus configs/soranexus/taira/validator_secrets.example.toml with python3 scripts/render_taira_validator_bundle.py --roster ... --secrets ... --output-dir ... before cutting traffic again.