fetch

name	fetch
description	Fetch content from URLs or file paths, auto-detect type, and convert to markdown
user-invocable	true
allowed-tools	Read, Write, Glob, Grep, Bash

/fetch

Fetch content from one or more URLs or file paths, auto-detect the source type, and convert to markdown.

Usage

/fetch <url-or-path> [url2 ...] [-o output_dir]

Workflow

1. Parse arguments. Extract -o <dir> if present (default: output/).
2. For each source URL or path: a. Run: bash fetch/ingest.sh "<source>" -o "<output_dir>"
   • ingest.sh auto-detects the source type (Google Drive, Google Docs, PDF, video, web) and routes to the appropriate parser. b. Check exit code. If the script fails:
   • Read stderr for the error message.
   • If a tool is missing, tell the user the install command.
   • If auth is needed (Google Drive), tell user to run gcloud auth application-default login.
   • If the conversion failed, suggest retrying with a different adapter. c. If successful, note the output file path (last line of stdout).
3. Report results: list all output files for successful fetches, and errors for failures.

Notes

• Do NOT clean up or enrich the content — just fetch and convert.
• Do NOT modify frontmatter beyond what the script generates.
• The output directory can point directly to another project's directory (e.g., -o ~/code/llm-wiki/raw/).
• Google Drive URLs are auto-detected: file type is determined from Drive API metadata.