FerroxLabs Build a KB Article
Estimated time: 20-30 minutes
This workflow turns a representative support ticket and its working resolution into a self-serve knowledge-base article plus a troubleshooting tree. It is interactive: the user reviews and refines the article, then reviews and refines the tree before shipping the package.
Steps
Step 1: Kick Off and Capture the Source Ticket (uses: support-kb-article)
Ask the user to paste a representative ticket for the recurring issue plus the resolution that actually worked. Capture both verbatim as the source ticket. Do not proceed until you have both the issue and the working resolution.
- Input: user-pasted representative ticket and working resolution
- Output: source ticket
- Key focus: capture both the problem and the proven fix before writing anything
Step 2: Infer Search Intent (uses: support-kb-article)
Using the support-kb-article skill, infer the search intent: the actual phrases users will type when they hit this problem. State the inferred search intents and the reasoning. Note that the article title and headers should match these intents. Capture the search intent.
- Input: source ticket from Step 1
- Output: search intent (phrases users will search)
- Key focus: derive real search phrasing so the article is findable
Step 3: Build the Article (uses: support-kb-article)
Using the support-kb-article skill with the source ticket and the search intent, draft the KB article: a title and headers matched to search intent, the issue description, and the step-by-step solution. Show the full article to the user.
- Input: source ticket from Step 1, search intent from Step 2
- Output: KB article draft
- Key focus: title and headers must match the inferred search intents
Step 4: Review the Article (uses: support-kb-article)
Present the article and ask the user whether to refine it or proceed to the troubleshooting tree. If the user wants changes, gather specific refinement feedback and re-run Step 3 with that feedback plus the prior article and the source ticket, looping up to 3 times until the user is satisfied. When satisfied, advance.
- Input: article draft from Step 3, user decision and refinement feedback
- Output: approved article
- Key focus: loop on the article only until the user approves (max 3 iterations)
Step 5: Build the Troubleshooting Tree (uses: support-kb-article)
Using the support-kb-article skill in troubleshooting-tree mode with the approved article, produce an if-this-then-that troubleshooting tree that branches on the symptoms a user might observe. Show the full tree to the user.
- Input: approved article from Step 4
- Output: troubleshooting tree (if-this-then-that branches)
- Key focus: branch on observable symptoms, each branch ending in a concrete action
Step 6: Final Review and Ship the Package (uses: support-kb-article)
Present the troubleshooting tree and ask the user whether to ship the KB article or refine the tree. If they want changes, gather feedback and re-run Step 5 with that feedback plus the approved article, looping up to 2 times. When approved, assemble the KB article package with these sections: Source Ticket, Search Intent, KB Article, Troubleshooting Tree.
- Input: troubleshooting tree from Step 5, approved article from Step 4, user decision
- Output: KB article package (assembled)
- Key focus: assemble the four-section package only after the user ships