Agentseek Lifecycle

AgentSeek Lifecycle

Core Workflow

1. Identify the workspace shape:
   • Generated project: look for .agentseek/lifecycle.toml, .env.example, pyproject.toml, frontend files, and project README.
   • Template project: look for cookiecutter.json, rendered {{ cookiecutter... }} paths, generated lifecycle specs, and template README.
2. Work through the public lifecycle surface:
   • agentseek create
   • agentseek doctor
   • agentseek dev
   • agentseek info
   • agentseek task
3. Treat .agentseek/lifecycle.toml as the compatibility contract between a generated project and AgentSeek.
4. Keep changes narrow:
   • Update generated instructions when lifecycle behavior changes.
   • Preserve non-interactive template rendering.
   • Use AgentSeek commands and project files as the validation surface.
5. Validate with public commands whenever possible.

Load References

• For lifecycle specs, task expectations, and public validation commands, read references/agentseek-lifecycle.md.
• For template variables, rendering, --no-input, private variables, hooks, and template maintenance, read references/cookiecutter.md.

Default Validation

For a generated project:

agentseek info
agentseek doctor
agentseek dev --dry-run
agentseek task --list

For a template project, render with defaults into a temporary directory, then run the generated project's lifecycle checks:

tmpdir="$(mktemp -d)"
cd "$tmpdir"
agentseek create <type>/<name> --no-input
cd <generated-project>
agentseek info
agentseek doctor
agentseek dev --dry-run

If the template lives on a branch that is not the template repository's default branch, pass the branch explicitly:

agentseek create <type>/<name> --checkout <branch> --no-input

If the repository provides its own test or lint commands, run the smallest relevant set after the public lifecycle check.