Install the skills

The get-nick-done package contains two halves: the gnd Python runtime and a set of agent skills. The skills are folders with a SKILL.md in the open Agent Skills format. Claude Code reads them from .claude/skills/ and Codex reads them from .codex/skills/. gnd install copies them there.

Install the runtime first from PyPI. Then choose direct skill files (this page) or the Claude Code or Codex marketplace plugin for each agent.

Run these commands in an empty directory, or in the project where your agent works:

gnd install --project

gnd doctor --project

install places every packaged skill for both agents. doctor reports what is placed in both the project and user scopes, because user-level skills take precedence over project-level ones. It also reports stale, missing, modified or colliding copies. A current install says nothing about whether your model or application passes.

Every installed skill is a self-contained folder:

from pathlib import Path

for root in (Path(".claude/skills"), Path(".codex/skills")):
    skills = sorted(p.parent.name for p in root.glob("gnd-*/SKILL.md"))
    print(root, len(skills), "skills:", ", ".join(skills))
    assert skills and "gnd-model-from-text" in skills and "gnd-dev-session" in skills

Choose where to install

Flag	Effect
`--project`	`./.claude/skills` and `./.codex/skills`; commit them with the project
`--global` (default)	`~/.claude/skills` and `$CODEX_HOME/skills`
`--claude` / `--codex`	only one agent
`--sync`	refresh every root that already has an install manifest, for example after `uv tool upgrade get-nick-done`
`--symlink`	link instead of copy; development installs only
`--force`	replace unmanaged files that block a write; never touches files you edited

gnd uninstall removes only what GND placed and leaves your edits alone. Do not resolve a collision by overwriting a skill directory by hand. Contributors adopting the repository's tracked links follow the separate contributor setup.

Agents without native skill discovery

Any agent that can read files can use the skills: point it at the relevant SKILL.md, for example "Read .claude/skills/gnd-model-from-text/SKILL.md and follow it." Each skill names the artifacts it writes and the gnd command that checks them.

What each skill produces →