Knowledge Base Updation Skill

The Knowledge Base update skill is a Pylon KB manager, not just a writing aid. It loads live Pylon Knowledge Base articles, identifies the docs affected by a process change, edits them as structured blocks, previews the result, gets approval, and applies approved updates through the Pylon API. It can also create new articles.

Download:

• FreightTiger Knowledge Base update skill

Why It Was Created

FreightTiger support processes, SLAs, troubleshooting steps, and SOPs change over time. When one process changes, the same information can exist in multiple Pylon docs. Manual updates are easy to miss and can leave agents or customers reading stale instructions.

The skill gives FreightTiger a controlled way to keep the KB consistent:

• Update every affected Pylon article after a process or SOP change.
• Create missing articles when a new workflow needs documentation.
• Move articles between collections.
• Change article visibility when content should become public, customer-only, internal-only, or AI-agent-only.
• Restore from local backups if an update needs to be rolled back.

User Story

When FreightTiger changes a support process, an operator can ask an AI assistant to update the Pylon KB. The assistant loads the current KB, lists the affected articles, proposes exact before/after changes, gets approval per article, then patches the approved articles in Pylon.

Example:

The Tier 1 acknowledgement SLA changed from 6 hours to 4 hours.
Update every Pylon KB article that mentions the old SLA.

How It Works

The package contains:

File	Purpose
SKILL.md	Operating instructions for the AI assistant or human operator.
USAGE.md	Installation and day-to-day usage guide.
BLOCKS.md	Article block schema for paragraphs, headings, lists, tables, code, horizontal rules, and preserved raw HTML.
pylon_kb.py	Python helper that calls the Pylon API, parses Pylon HTML into blocks, validates blocks, renders blocks back to Pylon HTML, applies updates, creates articles, and saves backups.

The helper supports:

• init: verify API access and return workspace/KB information.
• list-collections: list Pylon KB collections.
• list-articles --with-content: load articles with readable text.
• article-blocks: fetch one article as structured JSON blocks.
• preview-blocks: validate and preview proposed block JSON.
• patch-from-blocks: backup and update an existing article.
• create-from-blocks: create a new article.
• move-article: move an article to another collection.
• set-visibility: change article visibility.
• restore: restore an article from a backup HTML file.

The workflow is approval-gated. The assistant should not write to Pylon until the operator approves the affected article list, approves each article change, and gives final approval to apply.

Model-Agnostic Usage

The package is not tied to Codex or any single model.

For an AI agent with shell and file access, give it the extracted package and tell it to follow SKILL.md.

For a chat-only model, paste SKILL.md and BLOCKS.md into the conversation. A human operator runs the pylon_kb.py commands locally, pastes safe command outputs back to the model, saves the model's JSON block edits, previews them, and applies approved changes.

Do not paste Pylon API keys or .env contents into any model.

Setup

1. Download and extract the archive.

   mkdir -p ~/freighttiger-tools
   tar -xzf freighttiger-knowledge-base-update-skill.tar.gz -C ~/freighttiger-tools
   cd ~/freighttiger-tools/freighttiger-knowledge-base-update

2. Create a .env file in the package directory or a parent directory.

   PYLON_API_KEY=pylon_api_your_key_here

3. Add author user ID if you need to create new articles.

   PYLON_AUTHOR_USER_ID=01234567-89ab-cdef-0123-456789abcdef

4. Verify Pylon access.

   python3 pylon_kb.py init

   The response should include workspace and knowledge base details.

Update Existing Articles

1. Load collections and articles.

   python3 pylon_kb.py list-collections
   python3 pylon_kb.py list-articles --with-content

2. Ask the model to identify affected articles from the live article text.

3. Review and approve the affected article list.

4. Fetch each approved article as blocks.

   python3 pylon_kb.py article-blocks <article_id> > /tmp/pylon_<article_id>_original.json

5. Have the model edit only the required blocks and save a new block JSON file.

6. Preview the proposed result.

   python3 pylon_kb.py preview-blocks /tmp/pylon_<article_id>_new.json

7. Review the before/after diff and approve, skip, or refine.

8. After final approval, patch the article.

   python3 pylon_kb.py patch-from-blocks <article_id> /tmp/pylon_<article_id>_new.json

The helper saves a backup under /tmp/pylon_backups/ before patching.

Create A New Article

1. List collections and pick the destination collection.

   python3 pylon_kb.py list-collections

2. Confirm title, audience, visibility, and source material.

3. Create a block JSON file for the article.

4. Preview it.

   python3 pylon_kb.py preview-blocks /tmp/pylon_new_article.json

5. After approval, create the article.

   python3 pylon_kb.py create-from-blocks <collection_id> "<title>" /tmp/pylon_new_article.json

New article creation requires PYLON_AUTHOR_USER_ID.

Important Limits

• The helper works with published articles exposed by the Pylon API. Unpublished drafts may need to be handled in Pylon first.
• It does not detect concurrent edits made in the Pylon UI after the article is fetched.
• Backups are saved under /tmp/pylon_backups/; copy them somewhere permanent if long-term rollback is required.
• The operator must approve writes. The model should not decide to patch Pylon on its own.