Section 29 of 102
28. Reasoning, Planning and Recommendation
Stable section ID: S05-CON-010-SECTION-29 · 17 content blocks
Agent reasoning shall transform authorized information into explanations, alternatives, plans, or proposed actions while remaining within the agent’s defined role and authority.
Reasoning processes should consider:
- Applicable requirements.
- Current configuration.
- Physical state.
- Compatibility.
- Safety constraints.
- Lifecycle consequences.
- Cost and resource effects.
- Uncertainty.
- Available evidence.
- Human preferences.
- Reversibility.
- Required approvals.
Plans shall identify dependencies, assumptions, required tools, expected outcomes, verification steps, failure conditions, and recovery procedures.
Recommendations affecting safety, cost, configuration, or lifecycle performance should explain their material basis in a form suitable for the intended user. The explanation need not disclose proprietary internal model details, but it shall provide sufficient information for responsible evaluation.
An agent shall not present the existence of a plan as evidence that the plan is valid, compliant, or authorized.