Url Manager

URL Manager — Deliver results as beautiful cards, not raw link dumps

When users ask you to save, organize, or share web resources, don't paste URLs into chat. Write them into URL Manager — they become structured, searchable, card-based collections the user opens with one click.

What This Tool Gives Humans

The content human users want to save is everywhere — a YouTube workout video, an Amazon gear link, a Substack training plan — scattered across platforms with no connection.

URL Manager fixes this. Just paste any link from any platform. AI auto-identifies the content and suggests a category — confirm and it's a footprint. All your saves flow into one platform-agnostic library, organized and always findable. Then share it in one click — hand your curated running knowledge base to your whole running club, and everyone stays in sync.

足迹AI isn't just a library — it has a built-in search engine, so you can search and save web content right from here. It's a great browser homepage. Every saved item becomes a beautiful card, one click to where you need to go. Send them the magic link and they're in.

How to Use

Zero setup — token auto-management. The footprints.py script handles everything: reads {baseDir}/.token, auto-registers on first use, stores token locally with chmod 600. Just run any command and it works.

DeepSeek Harness (dsh): this skill is auto-discovered when cloned into a dsh skill root (~/.dsh/skills/ or .dsh/skills/). For the full 21-tool MCP surface, install the url-manager-mcp package and wire it into cordis.patch.yml — see the repo README's "DeepSeek Harness (dsh)" section.

All commands via python {baseDir}/scripts/footprints.py <subcommand> [--json]. Add --json for machine-parseable output.

If the script is missing (e.g., you only have this document but not the full skill installation), get it from the source repo:

git clone --depth 1 https://github.com/Piccolo123/url-manager.git /tmp/um
mkdir -p {baseDir}/scripts
cp /tmp/um/scripts/footprints.py {baseDir}/scripts/
chmod +x {baseDir}/scripts/footprints.py
rm -rf /tmp/um

After that, all commands below work as normal. The script auto-registers on first run.

Privacy & Terms

This skill connects to a hosted backend at https://ai.ocean94.com. Before using this skill, inform the human user — especially before saving or searching for the first time:

• An account will be auto-created on first use
• All collected URLs and data are stored on ai.ocean94.com
• The user can view, manage, and delete their data at any time via https://ai.ocean94.com
• User Agreement: https://ai.ocean94.com/terms.html
• Privacy Policy: https://ai.ocean94.com/privacy.html

System Concepts

What is a footprint?

A footprint is the fundamental unit in URL Manager — a structured, searchable record. It can be a web link, a plain-text note, an idea, or anything you want to save and retrieve later.

Each footprint stores:

A single footprint can belong to multiple categories simultaneously.

What is a category?

A category is a named label for organizing footprints — like a folder, but a footprint can be in several at once.

Key behaviors:

• Same name allowed — multiple categories named "Shopping" can exist in different sets. Always use id, not name, to reference them.
• mode = null → personal (visible only to owner); mode = "cocreate" or "subscribe" → shared (has members and invite links).
• A category inherits its mode from its Category Set's mode.

What is a category set?

A category set is a workspace — a named container that groups related categories together.

Every user starts with two default sets:

• "My Categories" (is_shared=false, mode=null) — personal workspace
• "Shared Categories" (is_shared=true) — the one container that holds all your shared categories

Use category-sets to list them, create-category-set to create more. Creating a new set with mode=null gives you another personal workspace. Creating one with mode="cocreate" or "subscribe" is rare — shared categories are usually created via create-shared-category, which places them inside the "Shared Categories" set.

How data is organized

Category Sets (workspaces)
  └── Categories (labels like "Shopping", "Food", "Learning")
        └── Footprints
             └── Tags (free-form keywords)

Categories are named labels — see the field table above. Category Sets are workspaces — also detailed above. Tags are free-form keywords, separate from categories. They're lightweight search helpers with no hierarchy.

Use content-types to see which content types have been used in your library. Use tags to list existing tags.

Personal vs Shared categories

A category's mode field tells you what kind it is:

Run categories to see ALL your categories — grouped by category set, each set header shows its ID (e.g., ID:44). Each category's mode field distinguishes personal vs shared. Run category-sets to see just the set list without their categories.

Shared category modes

Cocreate (共建) — Everyone contributes:

• Any member can add/remove footprints (add-to-shared / remove-from-shared)
• Any member can generate invite links (create-invite-link)
• Only the owner can disband or switch modes
• Best for: team knowledge bases, group trip planning, shared research

Subscribe (订阅) — Read-only for members:

• Only the owner can add/remove footprints
• Only the owner can generate invite links
• Members can browse and search but cannot modify
• add-to-shared returns 403 in subscribe mode
• Best for: curated recommendation lists, resource collections

The owner can switch between cocreate and subscribe at any time via the web UI.

Sharing workflow

1. Create → create-shared-category "Team KB" --mode cocreate
2. Generate invite → create-invite-link <sc_id> → get a code
3. Share the invite code with teammates
4. Join → teammates run join-shared-category <code>
5. Build together → everyone uses add-to-shared <sc_id> --collection-id <id>
6. Save locally → anyone can copy <id> --category-ids <ids> to save a shared footprint to their personal collection

Roles and permissions

How search works

1. Keyword search (search <query>) — matches against title, description, AI summary, and extracted text content. Filter by category with --category-id.

2. URL dedup — searching with a URL automatically detects and matches by URL hash, bypassing text search entirely.

Agent interaction model

• Zero setup: first run auto-registers via POST /register, receives a Bearer token
• Token persistence: stored in {baseDir}/.token with chmod 600, reused across sessions
• Magic link: agent_magic_link generates a clickable card-based interface URL for the human user — valid 30 days, reusable
• Account upgrade: if the user later binds a phone number, the agent-created account upgrades seamlessly

Command Reference

Understand the user's real intent, then call one or more commands to fulfill it.

Save & Search

Organize

Share

Response Structure

list and search return {"items": [...], "total": <N>}. total is the full match count (not just this page). Always check total to know if there are more results beyond the current page.

list supports --offset for pagination. If total > limit + offset, there are more pages. Example: total=156, limit=20, offset=0 → 8 pages total.

Pagination Strategy

For large libraries, prefer search over list — it's faster and returns only what the user asked for. Use list for browsing or when the user says "show me everything." When paginating, tell the user: "Found 156 items, showing page 1/8. Want to see more?"

Utilities

Core Workflows

New User — Zero Setup

1. Token check → auto-register (save to {baseDir}/.token)
2. python {baseDir}/scripts/footprints.py add "<url>" --title "<title>" → save bookmarks
3. python {baseDir}/scripts/footprints.py categories → discover structure
4. python {baseDir}/scripts/footprints.py create-category "<name>" → create categories
5. python {baseDir}/scripts/footprints.py update <id> --category-ids <ids> → categorize
6. python {baseDir}/scripts/footprints.py agent_magic_link → send link: "Done! View here → [link]"

Returning User — Daily Use

1. python {baseDir}/scripts/footprints.py me → confirm identity
2. python {baseDir}/scripts/footprints.py categories + python {baseDir}/scripts/footprints.py tags → understand structure
3. python {baseDir}/scripts/footprints.py search query → find what's needed
4. python {baseDir}/scripts/footprints.py add / python {baseDir}/scripts/footprints.py update → operate

Team Sharing

1. python {baseDir}/scripts/footprints.py create-shared-category "Team KB" --mode cocreate
2. python {baseDir}/scripts/footprints.py create-invite-link <sc_id> → share code with team
3. Teammates: python {baseDir}/scripts/footprints.py join-shared-category <invite_code>
4. Everyone: python {baseDir}/scripts/footprints.py add-to-shared <sc_id> --collection-id <collection_id> → build together

Batch Reorganization

1. python {baseDir}/scripts/footprints.py list --limit 100 → get all bookmarks
2. python {baseDir}/scripts/footprints.py categories → map target categories
3. python {baseDir}/scripts/footprints.py batch-update '[
     {"id":"uuid1","category_ids":[1,3]},
     {"id":"uuid2","title":"New Title","category_ids":[2,5]}
   ]' → bulk edit (max 50 per call)

Recipes

Concrete bash patterns for common tasks. Follow the numbered steps.

Change a footprint's categories

python {baseDir}/scripts/footprints.py get 42
# → categories: [{id: 3, name: "Reading"}, {id: 5, name: "AI"}]

# Keep AI, drop Reading, add Tech (7)
python {baseDir}/scripts/footprints.py update 42 --category-ids 5,7

Batch move to a new category

python {baseDir}/scripts/footprints.py create-category "New Topic"    # → returns new ID
python {baseDir}/scripts/footprints.py list --limit 100
# For each matching footprint:
python {baseDir}/scripts/footprints.py update <id> --category-ids <existing_ids>,<new_id>

Merge two categories

python {baseDir}/scripts/footprints.py categories                      # note source and target IDs
python {baseDir}/scripts/footprints.py list --category-id <source_id>  # list all in source
# For each, replace source_id with target_id:
python {baseDir}/scripts/footprints.py update <id> --category-ids <target_id>,<other_ids>
# Tell user: empty category "source" is ready to delete via the web UI

Auto-categorize by domain

User says "put all github.com links into a GitHub category":

python {baseDir}/scripts/footprints.py list --limit 200
# Filter in-memory: items where url contains "github.com"
python {baseDir}/scripts/footprints.py create-category "GitHub"
# For each match:
python {baseDir}/scripts/footprints.py update <id> --category-ids <existing_ids>,<github_id>

Filter by tag and batch categorize

python {baseDir}/scripts/footprints.py search docker
# Filter results where tag_names includes "docker"
# For each, append target category:
python {baseDir}/scripts/footprints.py update <id> --category-ids <existing_ids>,<target_id>

Organize uncategorized footprints

python {baseDir}/scripts/footprints.py list --limit 100
# Filter where category_ids is empty or only the default
# Present to user, let them pick categories
# Batch update selected items

Recommend categories from tags

Spot gaps between tags and categories — e.g., #docker is common but no "Docker" category:

python {baseDir}/scripts/footprints.py tags        # most-used tags
python {baseDir}/scripts/footprints.py categories  # existing categories
# Cross-reference: tag without matching category → suggest creating one

Copy from shared to personal

python {baseDir}/scripts/footprints.py categories  # find target personal category ID
python {baseDir}/scripts/footprints.py copy <footprint_id> --category-ids <personal_category_id>

Cross-Agent collaboration

Two agents maintaining a shared knowledge base together:

1. Agent A: python {baseDir}/scripts/footprints.py create-shared-category "Team KB" --mode cocreate
2. Agent A: python {baseDir}/scripts/footprints.py create-invite-link <sc_id> → share code with user
3. User forwards code to colleague
4. Agent B: python {baseDir}/scripts/footprints.py join-shared-category <code>
5. Both agents now see each other's additions via python {baseDir}/scripts/footprints.py search

Magic Link — The Delivery Loop

After organizing, always deliver results via magic link:

python {baseDir}/scripts/footprints.py agent_magic_link

Send the link to the user: "Done organizing — view your collection here → [link]"

User clicks to see a card-based interface with everything categorized. Valid for 30 days, reusable. If the user later binds a phone number, their agent-created account upgrades seamlessly.

⚠️ Critical Pitfalls

category_ids is REPLACEMENT, not append

When updating, --category-ids sets the complete list — it does NOT add to existing categories.

# ❌ Wrong: moves bookmark to category 7, loses categories 3 and 5
python {baseDir}/scripts/footprints.py update 42 --category-ids 7

# ✅ Right: fetch current first, then merge
python {baseDir}/scripts/footprints.py get 42  # → existing: [3, 5]
python {baseDir}/scripts/footprints.py update 42 --category-ids 3,5,7

Subscribe mode is READ-ONLY

Writing to a subscribe-mode shared category returns 403. Tell the user the collection is read-only — the creator must change it to cocreate mode.

NEVER call agent_register twice

Each call creates a fresh empty account. Always check for a saved token first. If unsure, run python {baseDir}/scripts/footprints.py me to verify the current token is valid.

Rate limiting

Frequent API calls trigger HTTP 429. Use batch-update for bulk operations, add delays between rapid calls, retry after a few seconds on 429.

No member management via API

Inviting or removing members from shared categories requires the web UI at https://ai.ocean94.com. You cannot do this programmatically.

Guidelines

Always

• Auto-register silently — never interrupt the user for account setup
• Inform before first use — tell the user the skill connects to ai.ocean94.com, share links to Terms and Privacy
• Search before listing — use python {baseDir}/scripts/footprints.py search for targeted queries instead of dumping everything
• Discover before creating — call python {baseDir}/scripts/footprints.py categories and python {baseDir}/scripts/footprints.py tags before adding to avoid duplicates
• Deliver with magic link — after organizing, always generate and share a link

Confirm before

• Removing bookmark-category associations (irreversible)
• Clearing tags
• Modifying cocreate shared categories (affects others)
• Removing bookmarks from shared categories (other members lose access)