Sync Ml Lab Docs

2. Update plugins/ml-lab/ML_LAB_FLOW.md

Extract the canonical mermaid flowchart from plugins/ml-lab/ml-lab.md. The flowchart is the primary visual specification — it must reflect:

• All nodes and their labels (especially PREFLIGHT, Gate 1, Gate 2, Gate 3, LOG)
• The LOG node caption (must match the log_entry.py invocation pattern in ml-lab.md)
• Gate labels (must match gate descriptions in ml-lab.md)
• Any new or removed nodes since the last sync

Rewrite the mermaid block in ML_LAB_FLOW.md to match.

3. Update README.md — mermaid block

Locate the mermaid flowchart block under the ### The Full Workflow section in README.md. Replace it with the same flowchart you just put in ML_LAB_FLOW.md. The block starts with ```mermaid and ends with ```.

4. Update README.md — prose sections derived from ml-lab.md

README.md contains several narrative sections that summarize or describe content from plugins/ml-lab/ml-lab.md. For each such section, compare against the corresponding source section in ml-lab.md and update the README to reflect the current authoritative content. Do not look for specific expected text — read both files and use judgment to identify any drift.

Sections to check include (but are not limited to):

• Any section describing the investigation logging workflow (schema, invocation, constraints)
• Any section describing gate behavior (Gate 1, Gate 2, Gate 3 — what triggers them, what blocks them, how they resolve)
• Any example run narrative or walkthrough prose that references workflow steps

For each section that has drifted from ml-lab.md: update the README to match the current behavior. Preserve README-appropriate voice (user-facing, explanatory) rather than copying agent-prompt text verbatim.

5. Report

List every change made, organized by file. For each change, give one line describing what was updated and why (e.g., "Updated LOG node caption to reference uv run log_entry.py"). If a file was already in sync, say so explicitly.

If no changes were needed anywhere, report "All downstream artifacts already in sync with plugins/ml-lab/ml-lab.md."

Do NOT remind the user to run claude plugin reinstall ml-lab. The PostToolUse hook (sync-plugin-cache.sh) fires automatically on every Edit or Write to any file under plugins/ml-lab/ and rsyncs the directory to the plugin cache immediately — manual reinstall is not needed.