0xenzyme/agent-harness/init

Harness Init

Use this skill only for explicit Harness setup, import, migration, or repair. Doctor and activation preview remain read-only.

The current user instruction and project instructions take precedence over this skill. This skill supplies routing and safety boundaries; it does not add an approval requirement beyond the boundaries stated below.

Read Setup And Migration Safety before any mutating init, import, or repair action.

1. Read repo instructions and inspect current state:

node <plugin-root>/scripts/agent-harness.mjs doctor --cwd <project>
node <plugin-root>/scripts/agent-harness.mjs config inspect --cwd <project> --json
node <plugin-root>/scripts/agent-harness.mjs config validate --cwd <project> --json

1. Before mutation, identify the existing Goal index and adapter paths. Never

create a second Goal index when importing an existing project.

1. Preview import/migration with --dry-run --json. Canonical output writes

contract, canonical paths, worktree.defaultPolicy, and completion gates. Legacy aliases are read for one migration boundary; conflicts fail.

1. Initialize or import only after explicit setup authority:

node <plugin-root>/scripts/agent-harness.mjs init --cwd <project> --contract adapter
node <plugin-root>/scripts/agent-harness.mjs config import --cwd <project> --task-index todolist.md --dry-run --json

1. Every planned write must pass project-root, configured-root, absolute/..,

and existing-parent realpath/symlink containment before any file is written.

1. Uninstall installed public skills only when the user asks to remove the

skill copies:

node <plugin-root>/scripts/agent-harness.mjs skills uninstall --cwd <project>

Preview with --dry-run --json. This removes recognized orient, intake, init, and execute copies under .agents/skills/ and byte-identical protocol copies under .agents/references/. It does not delete harness/ or .harness/ project data. Foreign skills and changed reference files stay in place.

Do not overwrite existing artifacts without explicit authority, edit AGENTS.md from an activation preview, start implementation, create workers, or perform delivery/deploy actions.