mcpField
S
support-deflection-knowledge-base-review

support-deflection-knowledge-base-review

github.com/sylphxaiOther139 installsUpdated today
Repository

About support-deflection-knowledge-base-review

This skill turns support deflection and knowledge-base questions into plans with owners, gates, metrics, and recovery paths. It is intended for teams reviewing help content, search, automation, and escalation, and specifies structured deliverables such as decision tables and QA checklists.

Key features

  1. 01Map contact drivers and user risks
  2. 02Classify support issues
  3. 03Define content and search strategy
  4. 04Set escalation and measurement rules
  5. 05Generate decision tables and QA checklists

Use cases

  1. 01Review help articles and search quality
  2. 02Set chatbot handling boundaries
  3. 03Plan escalation for unresolved users

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/sylphxai/skills --skill support-deflection-knowledge-base-review

After installing, type/support-deflection-knowledge-base-reviewin Claude Code to confirm it loaded.

Bundled filesagents/openai.yamlreferences/support-deflection-knowledge-base-patterns.md

Details

From the project README

Support Deflection Knowledge Base Review

Use this skill to convert a support deflection, knowledge base quality, help search, chatbot/macros, escalation, and unresolved-user recovery question into a concrete artifact with owners, gates, metrics, and recovery paths.

Workflow

  1. Identify top contact drivers, user stakes, self-service surfaces, article owners, search quality, escalation promise, and risk of trapping users.
  2. Read references/support-deflection-knowledge-base-patterns.md.
  3. Classify deflection target: how-to, troubleshooting, billing/refund, account access, bug/incident, policy, integration, or product education.
  4. Define content IA, search terms, article quality, diagnostics, macro/chatbot boundaries, escalation triggers, measurement, and product feedback loop.
  5. Produce support deflection plan, state machine, decision table, event schema, article QA checklist, and escalation policy.

Guardrails

  • Do not celebrate deflection rate if unresolved users churn, complain, refund, or recontact.
  • Do not block human escalation for billing, account access, safety, privacy, payment, or incident issues.
  • Do not let AI/chatbot answers outrank verified policy or current incident status.
  • Do not create duplicate stale help articles without ownership and refresh cadence.

Output format

Support context:
Contact driver / user stakes / escalation promise:

Deflection plan:
| Area | Decision | Evidence | Risk | Owner |
| --- | --- | --- | --- | --- |

Articles, search, and automation:
- <item> -> <policy, metric, edge case, support note>

Escalation and feedback loop:
- <trigger> -> <action, communication, owner>

Comments

Markdown supported