Miller web research

Skill anortham/miller/skills/miller-web-research

Use when researching web pages with Miller: fetch token-optimized markdown via browser39, import it as web content, then search/read bounded windows through Miller's content corpus.From its SKILL.md

Install

npx -y skills add anortham/miller --skill miller-web-research

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.1 KB, as published. Nobody here has run it

Miller Web Research

Use this workflow when web content should become searchable/readable through Miller without writing files under the repo.

Prerequisite

Check for browser39 first:

command -v browser39

If it is missing, tell the user:

browser39 is required for Miller web research. Install it with: cargo install browser39

Fetch And Import

Fetch into a temp markdown file, then import that file into Miller as web content:

tmp="$(mktemp -t miller-web.XXXXXX.md)"
cmd="$(mktemp -t miller-web.XXXXXX.jsonl)"
out="$(mktemp -t miller-web.XXXXXX.out.jsonl)"
url="https://example.com/page"
printf '{"id":"1","action":"fetch","v":1,"seq":1,"url":"%s","options":{"selector":"article","max_tokens":12000,"strip_nav":true,"include_links":true,"include_images":false}}\n' "$url" > "$cmd"
browser39 batch "$cmd" --output "$out"
title="$(python3 - "$out" "$tmp" <<'PY'
import json, sys
source, target = sys.argv[1], sys.argv[2]
with open(source, 'r', encoding='utf-8') as f:
    row = json.loads(f.readline())
markdown = row.get('markdown') or ''
title = row.get('title') or row.get('url') or 'web page'
with open(target, 'w', encoding='utf-8') as f:
    f.write(markdown)
print(title)
PY
)"
miller content add-markdown "$tmp" --url "$url" --display-path "$title" --json

Search And Read

Search only imported web content:

miller content search "important phrase" --kind web

Read a bounded line window from a hit:

miller content read --source-id SOURCE_ID --line 120 --context-lines 10

Rules

Do not create or modify repo files for web imports.
Do not write pages under docs or any tracked directory.
Do not paste full page markdown into the conversation.
Use content search --kind web first, then content read with a small window.
If browser39 returns truncated content, fetch the next offset and append it to the same temp markdown before one import.