Skip to main content
Rules and Skills are the two levers that determine how Gradial agents behave in your environment. Rules define what good looks like — the standards every output must meet. Skills define what agents are capable of — the behaviors they bring to each task. Together, they are what separates a generic AI execution from one that reflects your organization’s specific standards, voice, and way of working.

​
Rules

A rule is a defined standard that Gradial’s agents check against on every task. Rules cover brand and editorial compliance, SEO requirements, legal and regulatory obligations, accessibility standards, and functional quality. Once defined, rules run automatically — on every piece of content, every time, without manual QA cycles or sampling.

​
Where rules live

Rules are managed in the Rules Manager, accessible from the left sidebar. Rules are organized at two levels:
LevelScopeExamples
Organization RulesApply across all workspacesEnterprise brand standards, legal disclaimers, WCAG compliance
Workspace RulesScoped to a specific team, region, or projectProduct terminology, regional regulatory requirements, channel-specific tone
This hierarchy means your universal standards are always enforced, while individual teams can layer their own requirements on top.

​
Adding rules

There are three ways to add rules to Gradial: manually in the Rules Manager, by uploading existing documentation, or by asking an agent to create one for you.
Navigate to Rules Manager in the left sidebar, select the appropriate scope (Organization or Workspace), and click New Rule.Each rule includes:
  • Title — An action-oriented name that describes what it checks (e.g., “Address the reader directly with ‘you’”)
  • Description — Context on why this rule exists and what it enforces
  • Example — A before/after showing correct vs. incorrect implementation
  • Should do / Should not do — Specific guidance structured for agent interpretation
Choose which agents enforce the rule (Planning, Integration-specific, or both), then save and activate.

​
Assigning rules to agents

When creating or editing a rule, you specify which agents enforce it:

Planning Agents

Agents that coordinate content workflows — thread management, campaign orchestration, batch operations, chat interactions. Assigning a rule to Planning Agents means it’s applied early, before content reaches authoring. Enabled by default for new rules.

Integration-Specific Agents

The authoring and QA agents tied to your connected integrations (AEM, Sitecore, Marketo, SFMC, custom MCP). Authoring agents apply the rule while content is being created; QA agents apply it during quality review.
Assigning a rule to both means authors receive guidance while writing and QA catches anything that slipped through — two checkpoints, same standard.

​
Types of rules

CategoryWhat it covers
Brand & editorialVoice and tone, approved terminology, trademark usage, messaging consistency
SEOMeta description requirements, heading hierarchy, keyword usage guidelines
ComplianceRequired disclaimers, regulated language, regional legal requirements
AccessibilityAlt text standards, color contrast, heading structure, ARIA requirements
FunctionalBroken link detection, missing asset references, metadata completeness

​
Skills

Skills are the configured capabilities that shape how agents approach their work — beyond what Rules check, Skills define how agents plan, reason, and execute. Where Rules are explicit standards (“this content must include X”), Skills are behavioral configurations (“when working on this type of task, approach it this way”). Skills are configured per workspace and reflect how your team operates: which patterns agents follow when building pages, how they handle ambiguous instructions, what level of autonomy they apply for different task types, and how they structure outputs to match your CMS’s expectations.
Skills configuration is managed collaboratively with your Gradial team during onboarding and evolves as agents become more familiar with your environment. Contact your Gradial contact to discuss adding or adjusting Skills for your workspace.

​
How Rules and Skills work together

Rules and Skills operate at different layers of agent behavior: 
Task arrives
     ↓
Skills shape how the agent plans and approaches the work
     ↓
Agent executes, guided by Skills and your Design System patterns
     ↓
Rules validate the output before it reaches a reviewer
     ↓
QA report surfaces any failures with explanation and remediation guidance
The result: agents that work the way your organization works, and outputs that meet the standards your organization requires — without manual checks at every step.

​
Best practices

Start with high-impact rules. Focus first on the standards that catch the most common or most consequential issues. You can add nuance later — but starting with the rules that matter most gets teams to a consistent quality baseline quickly. Use the upload path to seed your ruleset. If your organization has existing brand guidelines or compliance checklists, uploading them is faster than manual entry and surfaces the full scope of your standards in one pass. Assign rules to both Planning and Integration agents. Early guidance from Planning Agents shapes how work is framed; Integration Agent checks catch anything that slipped through. Both layers together are more reliable than either alone. Review rules regularly. As your content strategy evolves, your ruleset should too. Rules that were relevant at launch may need updating — and new regulatory or brand requirements need to be added when they come into effect. Use workspace rules for team-specific needs. Keep Organization-level rules focused on universal standards. Use Workspace rules for team-specific requirements like regional compliance language or product-specific terminology.
Rules and Skills define what the agent should do. To see how outputs are validated against those standards, see QA Reports and Accessibility Reports.