Menu
Import web content cleanly into Evidence/imports/ with clutter removed. Use when importing a URL or HTML file as evidence for the project.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
The exact naming and structure convention for memory branch notes. Use when creating, naming, or reorganizing notes in Memory/, or when checking whether a note follows the convention.
Handle imported and extracted material correctly. Use when placing content in Evidence/, importing web pages or docs, recording captures, or deciding what is and isn't canonical.
Write memory notes that are pleasant in Obsidian without requiring Obsidian. Optional enhancement skill. Use when the user has Obsidian open alongside their agent workflow and wants better vault navigation.
Infer a project's natural ontology from the actual repo. Use when determining what memory branches to create, how to name them, and how to decompose the project into focused domains.
Write high-quality, stable memory notes. Use when creating or updating notes in agent-knowledge/Memory/ and you want specific guidance on format, quality, and what to include.
Read on session start. Manages adaptive durable memory across conversations. Use when reading or writing curated memory under agent-knowledge/Memory/.
| name | clean-web-import |
| description | Import web content cleanly into Evidence/imports/ with clutter removed. Use when importing a URL or HTML file as evidence for the project. |
Reduces noisy web content before it lands in the knowledge vault.
Strips navigation, ads, footers, and boilerplate. Produces readable markdown
stored in Evidence/imports/ -- clearly marked non-canonical.
Do NOT use for:
Evidence/raw/ directly)bedrock clean-import <url-or-file> --project .
Options:
--project <dir> -- project root (default: cwd)--slug <name> -- override output filename (default: derived from URL/title)--output-dir <dir> -- override output directory (default: agent-knowledge/Evidence/imports/)--dry-run -- preview without writing--json -- machine-readable outputThe command:
Evidence/imports/<slug>.md with proper frontmatter---
note_type: evidence
source: https://docs.example.com/api/rate-limits
title: Rate Limits - Example API
imported: 2025-01-15T14:30:00Z
canonical: false
---
# Rate Limits - Example API
(cleaned markdown content)
| Element | Action |
|---|---|
<nav>, <header>, <footer> | Removed entirely |
<aside> | Removed |
<script>, <style> | Removed |
<button>, <form>, <select> | Removed |
| Cookie banners, overlays | Removed (best effort) |
| Duplicate whitespace | Collapsed |
| Element | Converted to |
|---|---|
<h1> - <h6> | # - ###### |
<p> | Paragraph |
<ul>, <li> | Markdown list |
<pre>, <code> | Fenced code block |
<blockquote> | > |
<a href> | [text](url) |
If you prefer to import manually:
agent-knowledge/Evidence/imports/<slug>.md---
note_type: evidence
source: <url>
imported: <YYYY-MM-DDThh:mm:ssZ>
canonical: false
---
After importing, do NOT copy content directly to Memory/. Follow the evidence promotion process:
2025-01-15 - Updated from imported API docsSee evidence-handling skill for full promotion rules.
Imported web content goes stale. Add a review reminder comment if the content is time-sensitive:
---
note_type: evidence
source: https://...
imported: 2025-01-15T14:30:00Z
canonical: false
review_by: 2025-04-15
---
When imported: is older than 90 days, treat the content as potentially outdated
before promoting any facts to Memory/.