Agent onboarding

Connect an agent to the documentation and recipes in the project’s DASCore installation.

Keywords

agents, skills, cli, environment

DASCore provides installed API documentation and workflow recipes through its CLI. Agents search the generated Markdown with their existing file tools. Install the optional dependencies in the project’s Python environment using its package manager:

python -m pip install "dascore[agents]"

For a uv-managed project, use uv add "dascore[agents]". Follow the project’s version constraints. Then inspect the selected installation:

python -m dascore doc
python -m dascore skills
python -m dascore skill dascore
python -m dascore skill make-inventory

The Skills navigation and CLI use one catalog. Membership in dascore/docs/skills.yml makes a recipe a skill; its title, description, and procedure remain in the original documentation page. Recipes keep their ordinary .qmd filenames. Keyword tags describe topics independently of skill membership.

Install the agent bootstrap

Export the small native SKILL.md into the skill location supported by your agent. Installing the Python package does not change agent configuration. Use the selected project’s Python executable in the following commands.

Personal skills use these locations:

Agent	Directory
Codex	~/.agents/skills/dascore/
Claude Code	~/.claude/skills/dascore/

Run this Python example in the selected environment, changing .agents to .claude for Claude Code. It works on Windows, macOS, and Linux, preserves the exported UTF-8 bytes, and checks that export succeeded before replacing an existing skill:

import subprocess
import sys
from pathlib import Path

content = subprocess.check_output(
    [sys.executable, "-m", "dascore", "skill", "--bootstrap"]
)
directory = Path.home() / ".agents" / "skills" / "dascore"
directory.mkdir(parents=True, exist_ok=True)
(directory / "SKILL.md").write_bytes(content)

Other agents that support Agent Skills can use the same exported file in their supported skill directory. Reload or restart the agent if the new skill is not visible.

The bootstrap tells the agent to identify the project’s environment, verify dascore doc provenance, and retain that environment’s absolute Python executable and documentation directory. It loads the current routing recipe and specialized skills through the CLI, and searches the reported directory using the agent’s existing file tools. It contains neither a fixed interpreter path nor a copied inventory procedure.

Install the bootstrap once per agent location. Recipes and API documentation follow the selected DASCore installation, so switching projects also requires selecting that project’s interpreter. Update the exported bootstrap when its installation or discovery instructions change.

Check the connection

Ask the agent to use DASCore to inspect a small dataset or prepare an inventory. Check that it reports the intended Python executable, DASCore version, and documentation directory, searches that directory for relevant APIs or recipes, and uses that same executable for subsequent CLI calls. A global dascore command can belong to another environment.

The documentation cache may be outside the project workspace, so the agent should use its reported absolute path when searching. The bootstrap and recipes guide the requested work; they do not authorize publishing data or changing unrelated project configuration.