Update to Bagel

{
  "description": "Use when the user wants a bagel data pipeline — reducing a log to windows around events (\"keep 30s around every hard brake\"), recurring snippets/GIFs/exports, or any tasks-and-gates automation over robot data.",
  "included_files": [],
  "name": "authoring-pipelines",
  "skill_md_contents": "---\nname: authoring-pipelines\ndescription: Use when the user wants a bagel data pipeline — reducing a log to windows around events (\"keep 30s around every hard brake\"), recurring snippets/GIFs/exports, or any tasks-and-gates automation over robot data.\n---\n\n# Authoring bagel pipelines\n\nThe bagel MCP server owns the authoring workflow, including the\nreduce-vs-snippet decision, window/debounce extraction, and the\npreview-before-run rule. Do not write pipeline YAML from memory:\n\n1. Call `run_poml_capability` with `poml_path=\"./src/agent/compose/pipeline.poml\"`.\n2. Follow it exactly. In particular: always call `preview_pipeline` and show the\n   user the summary (events found, data kept) BEFORE running anything.\n3. Use `list_pipeline_capabilities` for the exact task/gate module paths and\n   arguments — never guess them.\n4. Execute the approved config through the MCP tools (`run_pipeline`, or\n   `run_pipeline_batch` for many sources). The capability also mentions a host\n   CLI (`run.py`); that path is for users at a terminal in the repo, not for\n   plugin sessions — do not shell out to it.\n\nIf the bagel tools are missing, Bagel is not connected yet. The user must\nstart the Docker container for their data format (see references/formats.md\nfor the format → image → extra-args table) and, in Codex, connect it once\nwith `codex mcp add bagel --url http://localhost:8000/mcp`. Claude Code\ninstalls connect automatically.\n"
}