Metacognitive Guard

Struggle signals include hedging language, deflection, verbose responses without clear conclusions, and hesitation on architectural decisions or ambiguous requirements. Metacognitive Guard monitors these indicators to flag when reasoning capacity is strained by competing constraints, multi-domain synthesis, or edge-case analysis requiring escalation.

Escalate when facing high-complexity problems: architectural decisions with conflicting tradeoffs, ambiguous requirements needing interpretation, synthesis across multiple domains, or intricate edge cases. Deep-thinking agents handle these better than comfortable reasoning capacity allows, producing more structured solutions.

Escalation requires integration with a self-assessment checklist and the Task tool. When struggle signals appear, spawn a deep-think-partner with a detailed prompt describing the problem, constraints, and context. The escalation produces structured prompts guiding focused analysis beyond initial reasoning limits.

Yes. Metacognitive Guard includes metrics tracking struggle detection frequency and escalation outcomes. This data reveals which problem types consistently exceed reasoning capacity, helping optimize when and how to route work to deep-thinking agents.

Hedging alone doesn't always warrant escalation. Escalate when hedging combines with structural complexity: competing architectural constraints, requirement ambiguity requiring interpretation, cross-domain synthesis, or edge cases defying straightforward analysis. The Guard distinguishes between linguistic uncertainty and genuine reasoning strain.