Claude Code and Codex

Claude Code remains the default runtime. Codex support is documented per package; do not assume that installing a plugin in one runtime makes it available in the other, or that a manifest proves equivalent behavior.

Claude Code

Run the selected package’s installer, then restart Claude Code:

npx @windyroad/planning

Invoke a skill by its slash command:

/wr-planning:outcome-led-planning

Codex

The planning package supports an explicit Codex installation:

npx @windyroad/planning --runtime codex --scope user

Restart Codex and check its installed plugin and skill lists. Select the skill name exposed by that installation. The outcome-led planning invocation is:

$wr-planning:outcome-led-planning

For a package that documents both-runtime installation, --runtime both selects both. This is an individual-package option; the current suite meta-installer does not offer that runtime selection.

Optional GitHub tracking

The full GitHub coordination requirements use planning 0.3.0 or later, or ITIL 3.3.0 or later through its compatibility command. Follow GitHub planning and agent coordination. Conversation and local planning remain available without GitHub access.

Install required reviewers

Outcome-led planning requires architecture, JTBD and risk assessment. Planning bundles no reviewers or hooks and does not require ITIL. Use available independent reviewers, or install the relevant plugins for the runtime you will use, following each package’s documentation:

Verify that all three substantive reviews are available before starting dependent work. The skill reports a missing review capability and holds implementation if it cannot obtain a review.

Understand evidence limits

An installation can register a skill without proving its later behavior. A working skill command does not establish every hook or cross-plugin workflow. Read package-specific support statements and verify the journey you depend on.