Loading…
Use when deciding whether and where to escalate a support ticket — encodes the org's escalation thresholds and targets (Sev-1 to oncall, refunds over a limit to a manager, Enterprise churn to retention, legal/press to legal) the model can't guess from the ticket alone. Template: replace with your thresholds and targets.
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| data-exposed | ·→· | ⚠ Error | 262% | 0% |
| invoice | ·→· | ⚠ Error | 256% | 0% |
| third-contact-login | ·→· | ⚠ Error | 265% | 0% |
| darkmode | ·→· | ⚠ Error | 259% | 0% |
| ent-api-500 | ·→· | ⚠ Error | 266% | 0% |
“The skill is a benign template outlining a support escalation policy for an agent to follow. It contains no executable code, external links, or suspicious instructions.”
> Template skill. The rules below are a worked example (a fictional "Acme"). Fork this skill and replace them with your organization's real rules — the agent then applies your policy instead of guessing. The measured lift demonstrates the shape works.
Decide where each ticket goes; reply with exactly one target. Rules (most specific wins):
vulnerability). Escalate immediately.
to legal; do not respond directly.
contacting for the 3rd+ time about the same unresolved issue.
normally, no escalation.