Skip to content

Ship Codex skills in the skeleton and scaffold lifecycle #1196

Description

@mohidmakhdoomi

Problem

Codev ships packaged Codev skills for Claude under codev-skeleton/.claude/skills/, but Codex does not receive the same scaffolded skills experience. Adding a second skill tree naively would also create two independently maintained sources that can silently drift.

The scaffold lifecycle must install and update Codex skills without overwriting user-customized skills, and documentation must describe both provider locations accurately.

Goal

Give Codex the same packaged Codev skills experience as Claude through codev-skeleton/.codex/skills/ and the scaffold lifecycle, while preserving user customizations and preventing silent drift between provider skill sets.

Scope

  • Add codev-skeleton/.codex/skills/ containing the Codev skills appropriate for Codex, at parity with the shipped codev-skeleton/.claude/skills/ set unless a skill is intentionally provider-specific.
  • Extend codev init, codev adopt, and codev update so Codex skills are installed/updated with the same preservation rules as Claude skills (do not overwrite user-customized existing skills).
  • Avoid two independently drifting skill sources: either generate/mirror from one canonical source or add an automated parity check with an explicit exception mechanism.
  • Update user-facing docs/output so both .claude/skills/ and .codex/skills/ are accurately described.
  • Respect Codev's dual-tree/framework rules and four-tier runtime resolution; do not assume a fresh adopter already has repository-local framework files.

Acceptance criteria

  • Fresh init creates .codex/skills/ from codev-skeleton.
  • Adopt and update add missing Codex skills without overwriting customized existing skills.
  • The shipped Codex skill set matches the Claude skill set except for explicit, tested provider-specific exceptions.
  • Automated tests prevent .claude/skills/ and .codex/skills/ from silently drifting.
  • Tests cover fresh installation, adding missing skills, and preserving customized existing skills.
  • User-facing documentation and command output accurately describe both skill locations.
  • The implementation follows the dual-tree and four-tier framework-resolution rules for fresh projects.

Related work

Recommended protocol

PIR: the preservation/parity approach needs plan review, and the init/adopt/update lifecycle should be exercised before PR creation.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions