System-1 decision models (Jev/Laya/Kev-class) as the governance layer for an LLM agent context lifecycle: growing association graph over session segments, relevance-gated recall with Trace-as-State ordering, and plan pre-ranking — measured in solve rate, cache hit/miss tokens, cost and latency. DSH plugin + harness-agnostic proxy.
See the context a Session submits to its model in the DeepSeek Harness Web UI: the system prompt, the tool schemas and every message, drawn as a grid of token squares.
Generic HTTP/API client plugin for DeepSeek Harness (DSH): one http_request tool for any method, headers, JSON/text bodies, redirects, timeouts and truncation, over a zero-dependency Node fetch driver
dsh (DeepSeek Harness) file explorer: browse the active session's workspace and preview files (Markdown/image/PDF/text/binary) right inside the chat GUI — a standard Cordis bundle plugin (read-only /dsh-files host routes + a conversation.view file tab).
Per-session Node REPL terminal in the DSH web UI: a session-local floating terminal (ghostty-web) bridged over WebSocket to an in-process node:repl that has the live agent Cordis Context (agent.ctx) in scope.
What will your agent's context cost before it runs? Audit instruction files, skills, MCP tool schemas and memory per source — with duplication, always-on versus on-demand costs, and a budget you can enforce in CI. Reads files and nothing else.