URL support for RLM analysis mode

[marknutter]

Problem

RLM analysis mode only works with local filesystem paths. To analyze a GitHub repo or web documentation, you have to manually clone/download first.

Proposed Solution

Detect URL arguments in the analysis path and auto-fetch content before entering the existing scan → chunk → dispatch pipeline.

URL types to support

GitHub repos (https://github.com/user/repo)

• git clone --depth 1 to a temp directory
• Optionally support branch/tag: https://github.com/user/repo/tree/branch
• Analyze the cloned directory normally

Web pages/docs (https://docs.example.com/guide)

• Fetch HTML, convert to markdown (or plain text)
• Save to temp file, analyze as a document

Raw file URLs (https://raw.githubusercontent.com/...)

• Download to temp file
• Analyze as a single file

Changes needed

1. rlm/cli.py or new rlm/fetcher.py — URL detection and download logic
2. skill/SKILL.md — Update analysis mode Step 0 to detect URLs and route through fetch step
3. Temp directory cleanup after analysis completes

Usage

/rlm "understand the hook/lifecycle system" https://github.com/openclaw/openclaw
/rlm "find security issues" https://github.com/someuser/somerepo
/rlm "summarize the API" https://docs.example.com/api-reference

Everything downstream (scan, chunk, extract, dispatch) stays unchanged — the URL is just resolved to a local path before the pipeline starts.