agentleFS
SkillMedevSkillMedev/customer-support-copilot/skills/kb-article-writer/SKILL.md

kb-article-writer

Skill0 starsChanged 39 days agoMITView on GitHub

Turns a resolved support ticket into a clear, searchable knowledge-base or help center article. Use when someone asks "write a KB article from this resolved ticket", "turn these resolved tickets into help center articles", or "write an internal knowledge base article for our support agents" - converting ticket resolutions into self-service documentation that deflects future tickets. Do NOT use for writing user-facing how-to guides from scratch - use help-documentation instead.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every ticket that gets answered manually is a ticket that could have been self-served.

    line 10
  2. Distinctive:

    Create an article when a ticket has been answered by two or more different customers in the past 60 days, or when a support lead flags a …

    line 14
  3. Distinctive:

    Use this structure for every article:

    line 18
  4. Distinctive:

    Every procedural step contains exactly one action.

    line 62
  5. Distinctive:

    No internal jargon, ticket language, or agent-only terminology survives into the article.

    line 63
  6. Distinctive:

    Do not copy-paste the agent's reply verbatim.

    line 68
  7. Distinctive:

    Do not bury the answer - lead with the most common scenario, not the edge case.

    line 69

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Knowledge Base Article Writer

Every ticket that gets answered manually is a ticket that could have been self-served. A good knowledge-base article is written for the customer who is about to open a ticket, not for the agent who already knows the answer.

When to create an article

Create an article when a ticket has been answered by two or more different customers in the past 60 days, or when a support lead flags a topic as recurring. Do not create articles for one-off edge cases or deprecated features.

Article structure

Use this structure for every article:

  1. Title: a plain-language question or task the customer is trying to accomplish ('How to reset your password', 'Why was I charged twice'). Use the exact words a customer would type into a search box.
  2. Summary (2-3 sentences): what this article covers and who it is for.
  3. Prerequisites (if applicable): what the customer needs before starting.
  4. Steps or explanation: numbered steps for procedural articles, short paragraphs for conceptual ones. Each step has one action. Use plain language, no internal jargon.
  5. What to do if it does not work: one to three common failure modes and their fixes.
  6. Related articles: two to four links to directly related help content.

Worked example

The same resolved ticket ("customer saw two identical charges; second was an authorization hold") turned into an article, badly and well:

Bad:

Title: Resolving duplicate charge incidents Opening: In certain scenarios, the billing subsystem may generate a secondary transaction record when a pre-authorization is initiated prior to capture...

Good:

Title: Why was I charged twice? Opening: If you see two identical charges from us, the second one is almost always a temporary authorization hold, not a real charge. It disappears from your statement on its own, usually within 3-5 business days. Here's how to tell the difference and what to do if both charges stay.

The bad version is titled in the team's internal vocabulary ("duplicate charge incident"), opens with system internals, and buries the answer. The good version is titled in the customer's search words, answers the most common scenario in the first two sentences, and only then handles the rarer case.

Writing rules

Write in second person ('you', not 'the user'). Use active voice. Keep sentences under 20 words. Screenshots or short screen recordings reduce confusion for procedural articles - include them where possible. Avoid mentioning specific version numbers or dates unless the content is truly version-specific; these make articles go stale.

SEO and findability

The article title and first paragraph must contain the keywords the customer would naturally use. Think about synonyms: a customer might search 'cancel subscription', 'end plan', 'stop billing', or 'delete account' for the same workflow - address each variant either in the title or in the first two sentences.

Review and freshness

Set a review date of 90 days on every new article. Mark articles as 'needs review' when the underlying product or policy changes. Archive articles that have not received a visit in 180 days - they add noise to search results.

Deliverable

Produce one publish-ready article containing: a search-phrased title, a 2-3 sentence summary, prerequisites where applicable, numbered steps or a short explanation, a "what to do if it does not work" section covering one to three failure modes, two to four related-article links, and a 90-day review date.

Quality bar

  • The title uses words a customer would actually type into search, not internal terminology.
  • The answer to the most common scenario appears within the first two sentences of the body.
  • Every procedural step contains exactly one action.
  • No internal jargon, ticket language, or agent-only terminology survives into the article.
  • Synonym variants of the customer's search are covered in the title or the first two sentences.

What to avoid

  • Do not copy-paste the agent's reply verbatim; ticket language is written for one customer, not all customers.
  • Do not bury the answer - lead with the most common scenario, not the edge case.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.