Shopfloor Agent Boundary Assessment

Role: Industrial Cybersecurity and OT Safety Architect with fifteen years securing programmable logic controllers and autonomous manufacturing agents.

Context

• Manufacturing site and unit: {{plant_facility}}
• Target autonomy classification: {{agent_autonomy_level}}
• Industrial network segmentation: {{scada_network_zone}}
• Safe operating envelope parameters: {{safety_critical_thresholds}}
• Applicable safety and compliance baseline: {{regulatory_standard}}
• Emergency stop and fallback standard: {{failure_containment_protocol}}

Task

Produce an operational guardrail assessment report establishing hard programmatic boundaries and fallback protocols for autonomous agents interacting with shopfloor systems in {{plant_facility}}.

Method

1. Review the operational profile of {{plant_facility}} alongside {{agent_autonomy_level}} to categorize physical hazard exposure.
2. Map incoming agent commands against network firewalls and isolation points in {{scada_network_zone}}.
3. Translate {{safety_critical_thresholds}} into strict deterministic pre-execution validation checks.
4. Define prompt filtering, tool permission allowlists, and parameter sanity filters to prevent command injection.
5. Align safety checks with compliance criteria mandated under {{regulatory_standard}}.
6. Specify automated circuit breakers and physical interlocks matching {{failure_containment_protocol}}.
7. Structure a monitoring and logging strategy for all blocked, altered, or failed agent actions.

Constraints

• All safety limits MUST be expressed as hard deterministic checks rather than probabilistic model inferences.
• Agent actions MUST NOT bypass existing hardware safety interlocks or programmable logic safety states.
• Recommendations must be actionable by automation engineers without deep AI safety specialization.
• Focus strictly on plant floor safety, operational continuity, and telemetry integrity.

Output format

An executive-ready report structured as follows:

• Executive Summary (150-200 words)
• Threat Surface and Physical Risk Matrix (table with 4-6 assessed vectors)
• Deterministic Guardrail Specification (pre-execution filters, runtime constraints, post-execution checks)
• Fail-Safe and Fallback Architecture (trigger conditions, recovery steps)
• Compliance Verification Table (aligned to {{regulatory_standard}})

Self-review

• Ensure every parameter in {{safety_critical_thresholds}} has a corresponding programmatic guardrail.
• Verify that the emergency protocols in {{failure_containment_protocol}} can trigger autonomously within millisecond tolerances.
• Check that network zoning rules in {{scada_network_zone}} are explicitly respected in the boundary definitions.