| name | endpoint-admission-voice-actions |
| description | Admit Penny voice endpoints with explicit identity, role, and action limits before they can issue backend actions. Use when edge voice devices, endpoint profiles, or voice-initiated tool permissions change. |
Endpoint Admission Voice Actions
Use this skill for the endpoint-admission and voice-action-governance lane in the distributed Penny stack.
Workflow
- Identify the endpoint class, owner, and intended role.
- Verify the minimum admitted metadata before allowing voice actions.
- Bind allowed tool families and confirmation rules to the endpoint profile.
- Check for drifted identities, duplicated endpoints, or stale admission state.
- Keep local audit records for endpoint-issued actions.
Required Evidence
- endpoint identity and role metadata
- allowed tool families or action limits
- confirmation policy for higher-risk actions
- drift or duplication check result
Guardrails
- Unknown endpoints fail closed.
- Voice input is not blanket tool authority.
- Admission state must be explicit and reviewable.
- Do not let endpoint convenience bypass local policy.
Best-Fit Companion Skills
$distributed-voice-plane-governance$skill-enforcer$local-compute-usage$docs-alignment-lock
Scope Boundary
Use this skill only for endpoint admission and voice-action permissioning.
For speech-loop transport behavior itself, route through $distributed-voice-plane-governance.
