Ensure your information sources are current and flag outdated data to maintain response quality.
Keeping the Knowledge Base Up to Date
1. Purpose
An agent's knowledge base is only as useful as its freshness. Outdated information leads to incorrect answers, user frustration, and trust erosion. This module defines the protocol for maintaining accurate, current, and complete knowledge.
2. Update Trigger Classification
Trigger Type
Description
Priority
Action
Version Release
New software version published
Critical
Update within 24h
API Change
Endpoint modified or deprecated
Critical
Update immediately
Feature Addition
New capability added
High
Update within 48h
Bug Fix
Known issue resolved
High
Update affected articles
User Feedback
Correction reported by user
Medium
Verify and update
Scheduled Review
Periodic accuracy check
Low
Monthly cycle
3. Update Process
Step 1: Detect Staleness
Monitor upstream changelogs and release notes
Track user reports flagging inaccurate content
Compare article timestamps against latest release dates
Flag articles older than 90 days for review
Step 2: Verify Current State
Cross-reference with official documentation
Test procedures described in the article
Confirm parameter names, formats, and default values
Validate code examples against current API
Step 3: Apply Updates
Preserve article structure and formatting standards
Update only sections that changed
Add version annotations where applicable
Include migration notes if behavior changed
Step 4: Validate
Review updated content for internal consistency
Ensure no broken cross-references
Verify code examples compile/execute correctly
Check that metadata (description, title) still matches content
4. Staleness Detection Matrix
Content Age
Risk Level
Action Required
< 30 days
Low
No action
30–60 days
Medium
Quick review
60–90 days
High
Full verification
> 90 days
Critical
Mandatory rewrite check
5. Source Priority for Updates
Official documentation — Always the primary source
Release changelogs — For version-specific changes
GitHub issues/PRs — For bug fixes and feature context
Community forums — For common user problems (verify independently)
User feedback — Valuable but always cross-check
6. Handling Contradictory Information
When sources disagree:
Trust official docs over community sources
Trust newer sources over older ones
Test the actual behavior when possible
Document the discrepancy if unresolvable
Flag for human review
7. Version Annotation Format
When content is version-specific, annotate clearly:
Knowledge Combination and Logical Reasoning for Agents — How AI agents should combine multiple information sources through logical reasoning, avoid irrelevant details, and synthesize knowledge into coherent, accurate responses.