This prompt sets up the PowerShell claude-codex-consult bridge, including in Codex CLI and other
agent hosts. For C3 (Rust) in Claude Code, use https://xelth.com/C3/#install-c3 and the C3 README
at https://github.com/xelth-com/C3#readme instead; the commands below are for the PowerShell bridge.

I want to use the claude-codex-consult bridge: either TRY one consultation now, or INSTALL it for
good and wire my reviewers. Ask me which - and which host I am in (Claude Code, Codex CLI, Z Code,
Kimi Code, Qwen Code, OpenCode, Muse Code or a plain shell) when you cannot tell.

If the plugin is not installed yet, give me the install commands for my host (README "Install" of
github.com/xelth-com/claude-codex-consult - in Claude Code: /plugin marketplace add
xelth-com/claude-codex-consult, then /plugin install codex-consult@claude-codex-consult) and wait:
installing is my action, in my own terminal. Never install software yourself.

TRY: one consultation needs one signed-in reviewer - the Codex CLI with a ChatGPT plan (codex login)
is enough, no roster. Check it with codex-providers.ps1 -Short; write a one-page brief about
something in this repository (the consult-codex skill's brief template); run codex-consult.ps1
-Task try-<date> -Purpose checkpoint -Brief <file> -Prompt "<one-line ask>" -ReplyName try; show
me the reply file and the findings it filed; tell me that telemetry is on by default and how to
switch it off (CODEX_CONSULT_TELEMETRY=off).

INSTALL: wire my reviewers by following the setup-providers skill from step 0
(codex-consult:setup-providers; a host without skills prints it with the plugin's own script:
powershell -NoProfile -ExecutionPolicy Bypass -File "<plugin>/scripts/codex-consult.ps1"
-Explain providers - pwsh -NoProfile -File on macOS and Linux; the plugin directory is in this
session's start line or under "Install" in the repository README).

1. Check first, ask second: the shell, git, the Codex CLI and its login, then the preflight
   codex-providers.ps1 -Short. One line per result.
2. Ask me which of these I have (several are possible), skipping what the preflight already shows
   as available: a ChatGPT plan (Plus, Pro, Team, Enterprise); z.ai GLM Coding Plan; Xiaomi MiMo
   Token Plan; Google AI Pro or Ultra (the Antigravity CLI); BytePlus ModelArk Coding Plan; Kimi
   Code membership; Alibaba Model Studio Token Plan (Qwen); Meta Muse Code subscription; a Claude
   subscription or an Anthropic API key (Claude Code; a coding plan with an Anthropic-compatible endpoint - z.ai,
   MiMo, Kimi Code - can also run through Claude Code, setup-providers section 3g); another
   Responses-API provider or a pay-as-you-go key; none of them.
3. For each one I name, follow its section of the skill. Ask before installing anything or editing
   config.toml. A CLI install, a login or an API key is mine to do in my own terminal: give me the
   exact command or the exact variable NAME, wait for me, then verify. Never create, print, read
   back or paste a key.
4. Write the roster (ask me which reviewer comes first and which count as "weighty"), verify with
   codex-providers.ps1 and a -DryRun consultation, and tell me that telemetry is on by default and
   how to switch it off (CODEX_CONSULT_TELEMETRY=off).
5. Record what you wired in the project's state.md (variable names, never values) and tell me what
   is still unavailable and why.