Google Doc To Md

Google Doc to Markdown

Download Google Docs as local markdown files. The Google Drive API natively supports text/markdown as an export format, so no external conversion tools are needed — the only variable is which account can authenticate to reach the doc.

Source routing

gws only authenticates your personal Google account. If the doc lives in an account gws can't reach (e.g. a work account), route through the next rung instead of giving up. Try rungs in order; only read/run the next rung on fallthrough — don't front-load setup you may not need.

Rung 1 — gws (default, unchanged)

Use the existing workflow below. This is the primary path whenever gws can authenticate to the doc's account.

Fall through to rung 2 if gws auth status fails, or the export/metadata call 403s/404s for the doc (wrong account, not the auth-expired case — that's a gws auth login fix, not a routing fallthrough).

Rung 2 — gcloud ADC (optional, only if configured)

Same clean native text/markdown export as rung 1 — same server-side exporter — just authenticated via gcloud Application Default Credentials instead of gws. Use this when the doc's account has ADC set up but not gws (e.g. a work account you've done gcloud auth application-default login for).

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/adc-check.sh"   # fast preflight; exit 0 = configured
bash "$SKILL_DIR/scripts/adc-export.sh"  [output.md]

If adc-check.sh fails, it prints an actionable one-line reason. Only then read [references/adc-setup.md](references/adc-setup.md) for the full gcloud setup steps — don't load it up front. If ADC isn't configured and setting it up isn't worth it right now, fall through to rung 3.

> Escaping note (all rungs): Drive's native exporter also > backslash-escapes markdown-significant punctuation that appears as > literal text in the doc (Catch up\!, \#123, (Q2 2026\)) — verified > live 2026-07-16 on a real doc. It's valid CommonMark, but if the raw > markdown is for human editing, the same de-escaper works on any rung. Codex: > this path resolves under Claude Code; substitute the directory containing this > SKILL.md in SKILL_DIR="${CLAUDE_SKILL_DIR}"; python3 "$SKILL_DIR/scripts/deescape.py" in.md out.md. > Docs that round-tripped from markdown import export clean; docs with > hand-typed punctuation don't. The connector rung always needs it.

Rung 3 — claude.ai Google Drive connector (zero setup, needs de-escaping)

Use when neither gws nor ADC can reach the doc's account, and the mcp__claude_ai_Google_Drive__* tools are available in this session (e.g. a work account added only via the Claude.ai connector). Verified live (2026-07-14, see the smoke-test notes): the connector's markdown export is backslash-escaped (\#, \[, \*, …) — there is no clean-export mode to prefer, a de-escape pass is required.

1. Resolve the doc ID (same URL pattern as rung 1: the segment after

/document/d/ up to the next /, ?, or #).

1. Call mcp__claude_ai_Google_Drive__get_file_metadata with fileId to

confirm access and get the title.

1. Call mcp__claude_ai_Google_Drive__download_file_content with

fileId and exportMimeType: "text/markdown". Base64-decode the returned content (e.g. python3 -c "import sys,base64; sys.stdout.write(base64.b64decode(sys.argv[1]).decode('utf-8'))" "$B64" or write to a temp file and base64 -d).

1. De-escape it:

``bash # Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md. SKILL_DIR="${CLAUDE_SKILL_DIR}" python3 "$SKILL_DIR/scripts/deescape.py" TEMP_INPUT.md CLEANED.md ``

1. Write CLEANED.md's content to the output file (same filename-derivation

rules as rung 1).

Known limitation: a literal backslash in the source prose immediately followed by punctuation (e.g. C:\*.txt) is indistinguishable from a connector-introduced escape and will also get unescaped. Low risk for prose docs; call it out if the source doc is code-heavy or path-heavy.

If none of the three rungs can reach the doc, fail with a clear message naming which rungs were tried and why each failed — don't silently give up after rung 1.

Prerequisites (rung 1)

```!

Codex: this path resolves under Claude Code; substitute the directory containing this plugin.

PLUGINROOT="${CLAUDEPLUGINROOT}" bash "$PLUGINROOT/scripts/auth-preflight.sh"

## Task (rung 1)

Run the entrypoint script, passing all arguments through:

```bash
# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" $ARGUMENTS

If no arguments were provided, ask the user for the Google Doc URL or ID and optionally the output file path.

Script Details

Downloading a Google Doc

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_ID_OR_URL

With a custom output path:

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_ID_OR_URL ./output.md

With --title flag to use the doc's title as the filename:

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_ID_OR_URL --title

How It Works

1. Extracts the doc ID from a URL if a full URL is provided
2. Fetches the document title from Google Drive metadata
3. Exports the doc as markdown via gws drive files export with

mimeType: text/markdown (native Drive API support)

1. Writes the result to the output file

Extracting a Doc ID

The doc ID is the long string in a Google Docs URL: https://docs.google.com/document/d/DOC_ID_HERE/edit

Output Filename

When no output path is given:

1. If --title is set, derives the filename from the Google Doc title

(lowercased, spaces to hyphens, .md extension)

1. Otherwise defaults to .md in the current directory

Export Size Limit

Google limits exported content from files.export to 10 MB.

Working with Native Doc Tabs

List a doc's tabs (id, index, title — indented by nesting):

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_ID --list-tabs

Export a single tab as markdown (basic fidelity: headings, bold/italic, links, lists, tables):

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_ID out.md --tab "Tab Title"

Note: the default (no --tab) Drive export flattens ALL tabs into one markdown file with each tab's title as a heading — fine for single-tab docs, confusing for multi-tab ones. Use --list-tabs first when unsure.

Batch Downloads

When downloading multiple docs, run in parallel:

# Codex: this path resolves under Claude Code; substitute the directory containing this SKILL.md.
SKILL_DIR="${CLAUDE_SKILL_DIR}"
bash "$SKILL_DIR/scripts/download.sh" DOC_URL_1 ./doc1.md &
bash "$SKILL_DIR/scripts/download.sh" DOC_URL_2 ./doc2.md &
wait

Additional Resources

If the bundled scripts are unavailable, see [MANUAL.md](references/MANUAL.md) for the step-by-step manual workflow.

Troubleshooting

Auth expired: Run gws auth login to re-authenticate. Wrong account: Run gws auth status to check which account is active. Account mismatch: If the doc belongs to a different Google account, the script will tell you which account you're authenticated as and suggest sharing or switching accounts. Run gws auth status to check. Empty output: The doc may be empty or the export may have failed — check stderr for error messages.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jtsternberg
• Source: jtsternberg/claude-plugins
• License: MIT

Install and usage instructions live in the source repository linked above.