nix-flake-update

Update Nix flake inputs in persops. Use when asked to update flake.lock/inputs, update AI inputs (codex-cli-nix, claude-code-nix, llm-agents), or run full `nix flake update`, then verify with `nix fla

SKILL.md

GitHub

Nix Flake Update

Workflow

  1. Decide scope
    • AI update: run nix flake update codex-cli-nix claude-code-nix llm-agents
    • Full update: run nix flake update (default when user does not mention AI)
  2. Switch
    • Run make local after any update
  3. Handle Pi changelog once
    • Run pi --version 2>&1 once after switching.
    • Compare it with modules/ai/pi/settings.json lastChangelogVersion.
    • If different, bump lastChangelogVersion to the current Pi version, read installed Pi CHANGELOG.md, and summarize entries between the old and new versions in the handoff.
    • If already equal, say no Pi changelog news.
  4. Verify (full updates only)
    • Run nix flake check
    • If failure, quote exact error, fix only update-related fallout, rerun nix flake check
  5. Commit
    • Stage only update-related files (typically flake.lock, modules/ai/pi/settings.json when Pi changed, maybe flake.nix or other necessary fixes)
    • Use Conventional Commit, keep commit isolated to the update
    • Examples: chore(nix): update flake inputs or chore(nix): update ai inputs

Notes

  • Skip nix flake check when updating AI inputs (unless explicitly requested).
  • Bumping Pi lastChangelogVersion is intentional when Pi updates; do not leave repeated startup changelog prompts for the user.
  • Work from repo root.
  • Keep changes minimal; no unrelated refactors.