Skip to main content

PIPE

Constant PIPE 

Source
pub const PIPE: &str = "# zag pipe\n\nChain results from completed sessions into a new agent session.\n\n## Synopsis\n\n    zag pipe [options] <session_ids>... -- <prompt>\n    zag pipe --tag <TAG> -- <prompt>\n\n## Description\n\nCollects the final results from one or more completed sessions and feeds them as context into a new agent session along with a user-provided prompt. This is the primary primitive for building DAG-like workflows where one session\'s output feeds into the next.\n\nResults are wrapped in `<session-result>` XML tags with session ID prefixes so the receiving agent can distinguish between sources.\n\n## Arguments\n\n    session_ids    One or more session IDs to collect results from\n    prompt         The prompt to send with the collected context (after --)\n\n## Flags\n\n    --tag <TAG>              Collect results from all sessions with this tag\n    -p, --provider <NAME>    Provider for the new session\n    -m, --model <NAME>       Model for the new session\n    -o, --output <FORMAT>    Output format (text, json, json-pretty)\n    --json                   Request JSON output from the agent\n    -a, --auto-approve       Skip permission prompts\n    -s, --system-prompt      Custom system prompt\n    --add-dir <PATH>         Additional directories (repeatable)\n    --file <PATH>            Attach a file to the prompt (repeatable)\n    --env <KEY=VALUE>        Environment variable for subprocess (repeatable)\n    --mcp-config <CONFIG>    MCP server config: JSON string or path to a JSON file (Claude only)\n    --size <SIZE>            Ollama model size\n    --max-turns <N>          Maximum agentic turns\n    -r, --root <PATH>        Root directory\n    -q, --quiet              Suppress logging\n    --name <NAME>            Name for the new pipe session (for discovery)\n    --description <TEXT>     Short description of the new pipe session\n    --timeout <DURATION>     Timeout (e.g., 30s, 5m, 1h); kills the agent if exceeded\n    -w, --worktree [<NAME>]  Create a git worktree for the new pipe session\n    --sandbox [<NAME>]       Run the new pipe session inside a Docker sandbox\n    --context <SESSION_ID>   Prepend another session\'s last result as context\n\n## Examples\n\n    # Pipe one session\'s result into a new one\n    sid=$(zag spawn \"analyze auth module\")\n    zag wait $sid\n    zag pipe $sid -- \"summarize the analysis\"\n\n    # Pipe multiple sessions\n    zag pipe $sid1 $sid2 $sid3 -- \"synthesize these analyses\"\n\n    # Pipe by tag\n    zag pipe --tag batch -- \"create a unified report\"\n\n    # With explicit provider/model\n    zag pipe --tag batch -p claude -m opus -- \"synthesize findings\"\n\n    # Name the follow-up session and bound it with a timeout\n    zag pipe \"$sid\" --name followup --timeout 5m -- \"summarize\"\n\n    # Run the follow-up in a worktree, prepending context from another session\n    zag pipe \"$sid\" --worktree --context \"$other_sid\" -- \"apply the plan\"\n\n## See Also\n\n    zag man collect    Gather raw results without re-processing\n    zag man spawn      Launch background sessions\n    zag man exec       Non-interactive single prompt\n";