knowledge-eval

Standardize knowledge evaluation for Panopticon agents by constructing a consistent POST /v1/knowledge/resolve payload. This automatically applies agent_slug and risk_level defaults, eliminating ad-hoc client-side logic and ensuring uniform governance across multiple agents.

The POST /v1/knowledge/resolve payload requests knowledge evaluation from Mission Control. It scopes results using tags, source_type, retrieval_mode, ranking_profile, and thresholds, returning a normalized summary with top evidence and actionable constraints.

Configure agent_slug, API URLs, risk level, and result limits by setting environment variables. The skill automatically reads these environment variables to populate the POST /v1/knowledge/resolve request, ensuring consistent agent scenarios without manual parameter passing.

Filter knowledge resolve results by applying optional tags, source_type, retrieval_mode, ranking_profile, and thresholds. These flexible filters scope the evaluation output for common agent scenarios, returning both a normalized summary and the raw resolve response.

The normalized knowledge evaluation summary includes top evidence, actionable constraints, and recommendations alongside the raw resolve response. This consolidates Mission Control outputs, providing agents with uniform evaluation data and governance information.

Use a dedicated skill for Mission Control knowledge evaluation when orchestrating multiple agents. It centralizes POST /v1/knowledge/resolve calls, ensuring uniform requests, consistent summaries, and automated governance defaults rather than maintaining scattered ad-hoc logic.