Behavior and output
Detects Claude Code, Codex, Cursor, Gemini CLI and Hermes from their configuration directories (or takes —client), copies every skill in the pack to each client’s skills directory with SHA-256 verification (installing only missing files and files this command installed before, never a file it did not write), and writes the MCP server entry for the view that client should use from the client table: Claude Code and Hermes get /mcp/all because they defer large catalogs themselves, the others get the bounded /mcp view. JSON configurations are merged, the Codex TOML gets a table appended when absent, and clients whose configuration this command does not edit (Hermes, and Claude Code at user scope) get the exact snippet or command to run. Records a receipt at .agentlinkops/agent-setup.json. Stores no credential; sign in through each client’s OAuth flow. —dry-run prints the plan and writes nothing. Exit 1 when a conflict was left for the person.Arguments
—client ID (repeatable: claude-code, codex, cursor, gemini-cli, hermes); —scope project (default) or user; —origin https URL (default https://app.agentlinkops.com); —pack DIR; —dry-run; —json.Exit status
0 on success; 1 when a conflict was reported; 2 on usage or configuration failure or when no client was detected. Read configuration and file ownership before automating this command. Read local file contracts for artifact fields. Browse the local command index for related commands.Check the selected environment and workspace access before calling hosted operations. See capability status.