Skip to content

[codex] Add cross-agent repo skills#28

Draft
alay2shah wants to merge 1 commit into
mainfrom
agent-skills
Draft

[codex] Add cross-agent repo skills#28
alay2shah wants to merge 1 commit into
mainfrom
agent-skills

Conversation

@alay2shah

Copy link
Copy Markdown
Collaborator

Summary

Adds a cross-agent usage layer for leap-finetune without changing runtime code:

  • expands AGENTS.md with concise repo guidance, core commands, storage rules, and a skill-selection table
  • adds CLAUDE.md as a small Claude Code wrapper that imports AGENTS.md
  • adds six focused lft-* Agent Skills for training, evals, rewards, data, launch, and checkpoints
  • exposes the same skills to Claude Code via .claude/skills -> ../.agents/skills
  • updates .gitignore so local Claude state stays ignored while the shared skills symlink is tracked

Why

The repo should be easy for Codex and Claude Code to use as a local-first model customization workspace. Always-on instructions stay short, while workflow-specific procedures live in skills that agents load only when relevant.

Validation

  • git diff --check
  • verified .claude/skills resolves to the shared lft-* skills
  • verified local .claude/settings.local.json and .claude/worktrees/ remain ignored

No runtime tests were run because this PR only adds agent instructions and skill files.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant