# Pandadoc Export

> Use this skill whenever the user wants to migrate, export, back up, or download data out of a PandaDoc account — documents, templates, or contacts. Trigger on phrases like "export PandaDoc", "migrate off PandaDoc", "back up our PandaDoc", "download all PandaDoc templates/documents", or "get my data out of PandaDoc before cancelling", even if they don't name the tool. Drives the `pandadoc-export` CLI (this repo).

- Skill: `josevitorls/pandadoc-export` (Agent Skill)
- Install (CLI): `npx skillmds@latest add josevitorls/pandadoc-export`
- Raw SKILL.md: https://api.skillmd.com/api/skills/josevitorls/pandadoc-export/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: josevitorls (https://skillmd.com/u/josevitorls)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/josevitorls/pandadoc-export

---


## Goal

Drive the `pandadoc-export` CLI to export all data from the user's PandaDoc account: documents (PDF + JSON), templates (JSON, plus PDF + DOCX via a headless browser), and contacts (JSON), organized per workspace.

## How the tool is structured

There are **two layers**, and they authenticate differently — this matters for how you run them:

- **API layer** (`discover`, `keys`, `inventory`, `extract`): uses an **Org-Admin API key** from `.env`. Fully unattended.
- **Browser layer** (`templates`): exports template **PDF + DOCX**, which the API cannot produce. It needs a **logged-in web session**, so the user logs in once in a browser window; the session is then reused headlessly.

## Steps

### 1. Check configuration

Confirm `.env` exists and contains a real `PANDADOC_ADMIN_KEY` (an Org-Admin production API key — a regular workspace key cannot list workspaces or mint per-workspace keys). If missing or still a placeholder, copy the example and ask the user to fill it in. Do not proceed without it.

```bash
cat .env 2>/dev/null || cp .env.example .env
```

### 2. Run the API export

```bash
pandadoc-export all
```

Runs, in order: `discover` (list workspaces) → `keys` (generate/reuse a key per workspace; the admin's own workspace is auto-detected and its key reused) → `inventory` (count docs/templates/contacts) → `extract` (documents as PDF+JSON by status with folders replicated, templates as JSON, contacts as JSON).

Note: `all` by itself does **not** export template PDF/DOCX — that is the browser layer (step 3).

### 3. Export template PDF + DOCX (browser layer)

Two options:

- As part of everything: `pandadoc-export all --with-templates` (opens a browser for login first), or
- On its own:

```bash
pandadoc-export login        # opens a browser — user logs in to PandaDoc once
pandadoc-export templates    # headless from here on
```

When a browser window opens, tell the user:

> "A browser window has opened — please log in to your PandaDoc account. The session is saved and reused, so you only do this once."

Wait for the user to confirm they logged in. `templates` auto-recovers across passes if the browser crashes; if the session later expires it stops and prints `LOGIN NEEDED` — run `login` again, then re-run `templates`.

### 4. Run individual stages (optional)

| Goal | Command |
|---|---|
| List workspaces | `pandadoc-export discover` |
| Generate per-workspace keys | `pandadoc-export keys` |
| Count items per workspace | `pandadoc-export inventory` |
| Download documents + templates(JSON) + contacts | `pandadoc-export extract` |
| Skip certain document statuses | `pandadoc-export extract --exclude-status voided,draft` |
| Log in for the browser layer | `pandadoc-export login` |
| Export template PDF + DOCX | `pandadoc-export templates` |

### 5. Resume an interrupted run

Every command is idempotent: re-running skips files already on disk and never regenerates an existing workspace key. Just re-run the same command. Recoverable failures are logged to `state/errors.jsonl`.

### 6. Verify output

```bash
ls output/
```

Each workspace is a folder named `{workspace_name}__{workspace_id}/` containing `contacts/`, `templates/`, `documents/<status>/...`, plus `_workspace.json` and `_manifest.json` (expected vs. obtained counts). Template `.json`, `.pdf`, and `.docx` share the same filename stem so they sort together.