AgentAction
← All agent recipes

Customer support / Read only

Triage a support request

Required serversSupport + Knowledge

Recipe maintained by AgentActionVersion 1.0.0Evidence: synthetic fixtures

The job

Produce a policy-grounded support recommendation with clear evidence and an explicit escalation when facts are insufficient.

Connections you’ll need

Use servers that expose these tools. Review their schemas and permissions in your agent runtime before running the recipe.

Support

Read a bounded ticket and customer-provided context.

get_ticket

Knowledge

Retrieve the applicable support policy.

search_policyget_policy

Operating boundaries

  • Read only: do not change tickets, issue refunds or send messages.
  • Ask for missing facts; escalate ambiguous or conflicting policies.
  • Keep customer information inside the approved tools and the current response.
  • A correct escalation can satisfy the task; do not invent certainty.

What a completed job requires

  • A recommendation or explicit escalation is present
  • Applicable policy rules are cited
  • No mutating tool calls occurred

Inspect the test cases

Deterministic checks against supplied synthetic observations. No live model or connected service was tested by these checks.

Recommendation with policy citationsExpected: met
{
  "recommendation_present": true,
  "policy_cited": true,
  "mutating_calls": 0
}

Expected evaluation: met

  • A recommendation or explicit escalation is present: expects true
  • Applicable policy rules are cited: expects true
  • No mutating tool calls occurred: expects 0
Recommendation without a policy citationExpected: not met
{
  "recommendation_present": true,
  "policy_cited": false,
  "mutating_calls": 0
}

Expected evaluation: not_met

  • A recommendation or explicit escalation is present: expects true
  • Applicable policy rules are cited: expects true
  • No mutating tool calls occurred: expects 0
Unexpected ticket mutationExpected: not met
{
  "recommendation_present": true,
  "policy_cited": true,
  "mutating_calls": 1
}

Expected evaluation: not_met

  • A recommendation or explicit escalation is present: expects true
  • Applicable policy rules are cited: expects true
  • No mutating tool calls occurred: expects 0
Tool-call evidence unavailableExpected: indeterminate
{
  "recommendation_present": true,
  "policy_cited": true
}

Expected evaluation: indeterminate

  • A recommendation or explicit escalation is present: expects true
  • Applicable policy rules are cited: expects true
  • No mutating tool calls occurred: expects 0

The agent’s instructions

  1. Read the specified ticket and retrieve the applicable policy using the connected Support and Knowledge tools.
  2. State the recommendation, cite the relevant policy rules, and identify the facts supporting it.
  3. If the policy or ticket is ambiguous, recommend manual review and explain which information is missing.
  4. Do not call mutating tools. Return the recommendation and evidence to the user.
  5. For observability, report whether the recommendation is present, policy citations are present, and no mutating tool calls occurred. Distinguish your self-assessment from independently observed evidence.

Follow it into operation

Connect your agent to an AgentAction workspace. Jobs shows what happened on each run; Evals lets you configure evaluation definitions and routing. Review the provenance of each result: agent self-assessments and independent observations provide different evidence.

Downloading a recipe does not connect accounts, deploy an agent, install controls, or configure Evals.