# Confluence To Md > Convert Confluence pages to clean GitHub-Flavored Markdown files. Use this skill whenever the user wants to pull a Confluence page into the local repo, convert Confluence content to markdown, download a spec from Confluence, or batch-convert multiple pages. Also triggers on Confluence URLs, page IDs, or phrases like "pull this page", "download this spec", "confluence to md". Use this even if the user just pastes a Confluence URL and says something vague like "get this" or "convert this". - Skill: `qa-aman/confluence-to-md` (Agent Skill, multi-file: 7 files) - Install (CLI): `npx skillmds@latest add qa-aman/confluence-to-md` - Raw SKILL.md: https://api.skillmd.com/api/skills/qa-aman/confluence-to-md/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: qa-aman (https://skillmd.com/u/qa-aman) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/qa-aman/confluence-to-md --- # Confluence to Markdown Pull Confluence pages into the local repo as clean markdown. The bundled script handles all the messy Confluence storage format edge cases (user mentions, complex tables, images, status badges) so you don't have to. Zero LLM tokens spent on the actual conversion. ## Quick start ```bash # Single page python3 .claude/skills/confluence-to-md/scripts/confluence-to-md.py # Batch (write a Python loop - don't run CLI commands one by one) python3 scripts/batch-convert.py ``` The script auto-loads credentials from `.env` at the project root. ## Workflow ### Step 1: Get the page ID Extract from the Confluence URL - it's the number in the path: ``` https://[your-instance].atlassian.net/wiki/spaces/[SPACE]/pages/1234567890/Page+Title ^^^^^^^^^^ this ``` If the user gives a title instead of URL, search via API: ```bash curl -s -u "$CONFLUENCE_EMAIL:$CONFLUENCE_TOKEN" \ "$CONFLUENCE_BASE_URL/wiki/rest/api/content?spaceKey=[SPACE]&title=$(python3 -c 'import urllib.parse; print(urllib.parse.quote("Page Title"))')" ``` ### Step 2: Pick the output path Follow your project's naming conventions. A common pattern: | Content type | Path pattern | |---|---| | Feature specs | `outputs/features/NNN-feature-name/feature-name.md` | | Knowledge docs | `inputs/knowledge/slugified-title.md` | | Cross-functional | `outputs/cross-functional/doc-name.md` | **Folder naming** - strip from Confluence titles: 1. Author bracket tags: `[Author Name] SPEC | Feature` -> `feature` 2. Doc type prefixes: `SPEC |`, `PRD |`, `DOC |` -> remove 3. Phase numbers when a product name exists: `Phase 2 | Product Overview` -> use product name 4. Use the feature/product concept name, not the page description ### Step 3: Run the script ```bash python3 .claude/skills/confluence-to-md/scripts/confluence-to-md.py 1234567890 outputs/features/001-my-feature/my-feature.md ``` ### Step 4: Post-conversion check 1. Scan for raw HTML `` tags with `data-*` attributes (embedded media) - replace with markdown links 2. Verify image references work (script downloads to `images/` subfolder automatically) 3. Check for excessive blank lines or formatting artifacts ## Batch conversion Write a Python driver script for multiple pages: ```python import subprocess pages = { "1234567890": "outputs/000-topic-a/topic-a.md", "9876543210": "outputs/001-topic-b/topic-b.md", } for page_id, output_path in pages.items(): subprocess.run(["python3", ".claude/skills/confluence-to-md/scripts/confluence-to-md.py", page_id, output_path], check=True) ``` ## Index files (zero LLM tokens) For overview pages, don't convert - parse programmatically: 1. Fetch page HTML via REST API (`body.storage`) 2. Parse `` tags with regex 3. Resolve titles to URLs via API 4. Write structured markdown ## What the script handles The script automatically resolves these Confluence storage format quirks: - `