kata-agent-overview-design

To generate governance-grade agent identity documentation and a system prompt, use a workflow that drafts overview.md and system-prompt.md by applying stepwise sourcing from PoV, DoOC, or ADR history, and validates reciprocity against feature docs.

Reciprocity validation in agent design is the process of verifying that each served feature exists and declares served_by_agents, ensuring the agent overview aligns with feature-agent mapping before finalizing the system prompt.

Structuring a system prompt using the lex-system-prompt specification requires a consolidated four-block format, including a mandatory Block 1 stage declaration and OWASP LLM Top 10 2025 boundary guardrails integrated into Block 2.

Agent overview generation requires entry-mode-specific sourcing from PoV, DoOC, or ADR history, meaning you must provide a valid POV source path or ADR reference to produce the canonical overview.md and system-prompt.md outputs.

The best way to enforce boundary guardrails in an agent system prompt is to mandate inclusion of OWASP LLM Top 10 2025 security boundaries within Block 2 of the lex-system-prompt-aligned four-block structure during stepwise drafting.

Agent overview reciprocity validation fails when listed served feature slugs do not exist in the documentation or when those feature docs lack the required served_by_agents mapping pointing back to the target agent.