knowledge-rag

Use when designing or extending this project as an OpenCode-based agent framework with minimal custom bottom-layer work.

Use when helping a user understand, generate, review, or revise agent.yaml, choose skills, set knowledge and memory policies, define gates, configure models…

Use when answering questions, making recommendations, updating memory, proposing skill changes, or passing gates where claims must be grounded in citable…

Use when deciding whether short-term state, user feedback, repeated failures, decisions, preferences, or project facts should become long-term semantic…

Use when planning or performing code, config, skill, documentation, or memory edits where the agent must make the smallest evidence-backed modification, avoid…

Use when a user starts or revises a heavy-reasoning agent project and the agent must clarify goals, audience, scope, constraints, success criteria, failure…

Use when the agent should learn from failures, feedback, tests, or repeated patterns by generating reviewed proposals for skill, memory, config, or workflow…

Use when an agent action needs risk assessment, approval, state-machine routing, interrupt handling, or enforcement of evidence requirements before answering…

Use when defining, running, or reviewing tests and acceptance checks for a heavy-reasoning agent, especially to prove evidence binding, gate enforcement…