AugmentClaude

Defuddle

Extract web pages as clean, readable Markdown with clutter removed.

Installation

  1. Make sure Claude is on your device and in your terminal.

    Skills load from ~/.claude/skills/ when Claude Code starts up — so you need it on your machine first. If you don't have it yet, install it once with the command below, then run claude in any terminal to verify.

    One-time setup
    npm i -g @anthropic-ai/claude-code

    Already have it? Skip ahead.

  2. Paste into Claude Code or into your terminal.

    This copies the whole skill folder into ~/.claude/skills/defuddle-agricidaniel/ — the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Safe default: works for every skill.

    Faster alternative (instruction-only skills)

    Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates — they won't be downloaded and the skill will fail when it tries to load them.

    Quick install (SKILL.md only)
    Sign up to copy
  3. Restart Claude Code.

    Quit and reopen Claude Code (or any other agent that loads from ~/.claude/skills/). New skills are picked up on startup.

  4. Just ask Claude.

    Skills auto-activate when your request matches the skill's description — no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the “What this skill does” section above.

Prefer to read the source first? Open on GitHub.

When Claude uses it

Plan optional Defuddle extraction and retain already available local source files through reviewed filesystem capture. Use for defuddle, clean this URL, strip page clutter, readable Markdown from a web page, or preparing a web source for later wiki ingestion. Network extraction remains unavailable until an enforcing adapter is verified.

What this skill does

Defuddle

Treat Defuddle as an optional external extractor. This package supports inert URL plans and filesystem capture; it does not ship an enforcing network cleaner. Cleaning, raw capture, and wiki ingestion are separate operations.

Resolve the installed product root from this skill's own location, never from the selected vault or process working directory:

PRODUCT_ROOT=/absolute/path/to/installed/claude-obsidian
CORE="$PRODUCT_ROOT/scripts/claude-obsidian.py"
test -f "$CORE"

Every ../wiki/references/ link in this file resolves the same way, relative to this skill's own directory under $PRODUCT_ROOT, never relative to the selected vault's wiki/ directory. Upstream flags may be checked in a separately installed kepano/obsidian-skills reference, but do not override this package's privacy and transaction boundary.

Inert URL plan

Accept HTTPS URLs only. Reject credentials, fragments, controls, sensitive query parameters, and non-public hosts or IP addresses. Never interpolate a URL into a shell string. Use one argv element when constructing a plan:

python3 "$CORE" capture external-plan url "HTTPS_URL"
python3 "$CORE" contracts --verify --capability defuddle --vault "$VAULT"

The plan executes no network request. Report the normalized host, required network egress, redirect policy, optional dependency, and execute: false. An available state means no executable was found. A configured state means an executable was discovered without a bundled behavioral verifier. Identify the manual executable and its reported reason; never run it merely because it was found. Never relabel configured as verified.

Why execution stops here

The upstream CLI inspected at commit 556d74e8c8fd68c49912d4b11a4ae9ec04d7b2b5 (checked 2026-10-17) follows redirects and enables asynchronous extraction. Even parsing a local HTML file can request additional resources or third-party fallbacks. Inspecting argv or obtaining consent for the initial URL cannot prove that every eventual destination remains approved.

Do not execute the bare CLI against a URL or local HTML under this workflow. Do not silently substitute a browser, fetcher, proxy, cookies, or another runner. A future runnable adapter must enforce approved destinations after DNS and on each redirect, disable uncontrolled async/fallback traffic, constrain proxy behavior, and bound bytes, time, and subprocess output. It must retain original and final locators, exact version, approved hosts, capture hash, and mode. Behavioral tests must prove these limits before execution is offered. Consent for each destination remains required even with an enforcing adapter.

Do not install tools or claim cleaning, extraction quality, or token savings. Leave the URL queued, or accept an already available local file. A local file may be retained as bytes without rendering HTML or running its scripts.

Reviewable filesystem capture

When the user asks to retain an existing local HTML or Markdown file, place it in the selected vault's inbox/ first. Preview and apply the filesystem plan with a pinned timestamp and operation ID:

python3 "$CORE" capture plan --vault "$VAULT" "$VAULT/inbox/source.md"
python3 "$CORE" capture apply --vault "$VAULT" "$VAULT/inbox/source.md" \
  --generated-at "$GENERATED_AT" --operation-id "$OPERATION_ID"
# Set APPROVAL_SHA256 to the reviewed approved_plan_sha256 from the dry run.
python3 "$CORE" capture apply --vault "$VAULT" "$VAULT/inbox/source.md" \
  --generated-at "$GENERATED_AT" --operation-id "$OPERATION_ID" \
  --approved-plan-sha256 "$APPROVAL_SHA256" --apply

This preserves source bytes in create-only .raw/captured/ storage. It does not clean HTML, verify source metadata, or synthesize pages. Inspect the actual result and record extraction limitations. If already cleaned text is supplied, retain its known extractor/version and locator without inventing missing fields. Report the operation ID and exact changed paths. A repeated identical payload is a no-op; never overwrite immutable raw state.

Use operation-transactions.md for custom capture drafts. Read or analyze transiently when retention was not requested. Invoke wiki-ingest only as a distinct requested operation.

Related skills