forcedotcom/sf-skills / plugins/builder/service-engagement/skills/service-agentforce-human-escalation-configure
service-agentforce-human-escalation-configure
UnverifiedUse to configure and verify Agentforce agent-to-human escalation, including human handoff, a staffed fallback queue, and failure-threshold directives. Triggers: configure agent escalation, escalate to a human agent, agent handoff to a queue, configure a human escalation queue, set an Agentforce fallback handoff. Do not use to create an agent or configure inbound channel routing; use service-agentforce-channel-configure for inbound routing.
references/failure-policy.md
View on GitHub# Failure-threshold policy (native, directive-based)
## Why this is not a metadata counter
Agentforce does not expose a metadata field or setting that counts "consecutive failed
turns" and escalates automatically. The "escalate after two failures" behaviour is a
**planner policy expressed as instruction text** in the agent's Agent Script (NGA
authoring model). This is the platform-native representation the user selected.
Consequences:
- The threshold is **authored, not enforced by a counter**. The bundled template
[`assets/escalation-thresholds.instructions.md`](../assets/escalation-thresholds.instructions.md)
is the source directive text.
- It is **verified by an ADK eval rubric** (does the agent escalate on the 2nd failure? on
the 1st failure for password reset?) and a **documented manual conversation test** — never
by a deterministic metadata round-trip.
- The `verify-and-configure.sh` verdict therefore reports the threshold as `directive`
(authored / not-authored), distinct from the `deterministic` config surfaces
(`canEscalate`, `outboundRouteConfigs` name+type coupling on the Messaging surface, outbound
flow Active, queue + QueueSobject + bound QueueRoutingConfig, ≥1 active human member, agent Active).
## Recommended thresholds
The threshold is a **caller input**, not a fixed constant. The values below are defaults;
a scenario that mandates a different number (e.g. a narrative requiring escalation after
**three** consecutive failures) supplies its own value via the skill input — no code change
is needed.
| Scenario | Threshold | Rationale |
|---|---|---|
| Default (any unresolved request) | 2 consecutive failures | Give the agent one retry before handing off |
| Password reset / credential issues | 1 failure | Time-sensitive; do not retry repeatedly |
| Caller-specified | any N | Whatever the scenario's authoritative requirement is (pass N as the input) |
## Authoring steps
1. Substitute tokens in the template (`__DEFAULT_FAILURES__=2`, `__RESET_TOPIC_LABEL__`,
`__ESCALATION_MESSAGE__`).
2. Add the default directive to the agent's global/system instructions.
3. Add the per-topic override to the password-reset topic instructions.
4. Republish + reactivate the agent (`sf agent validate|publish|activate ... --target-org $ORG`).
5. Confirm via the eval rubric and the manual test in
[runtime-verification.md](./runtime-verification.md).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |