A Programming Paradigm for Spatiotemporal Composability
Modern software—from plugin systems to self-evolving agent harnesses—increasingly requires dynamic composition, yet its formal foundations remain underdeveloped. A future harness
River
Primary-source items first added here within the past 48 hours.
@deepseek-ai/dsh-user-approval
Channel-neutral one-shot approval seam. ctx.approval.request(req) returns allowed-once, rejected, cancelled, or unavailable; missing or failing answerers fail closed, and
dsh-sandbox-policy
mode — the deployment default SandboxMode (read-only / workspace-write / danger-full-access), validated at load. Default read-only (fail-safe). Known limitation: Fil
@deepseek-ai/dsh-permission-presets
Each configured name bundles sandbox/mode with approval/policy; the defaults are workspace-write (workspace-write + ask) and danger-full-access (danger-full-access +
Self-referential Cordis toolset
The self-referential Cordis toolset: five model-facing tools over the live runtime in the current DSH process. The sandbox isolates globals but is not a security boundary. Node gl
DeepSeek Harness README at b150a55
DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI. It uses an architecture where everything is a plugin, and is powered by Cordis, whose design
Minimal preset at b150a55
The minimal agent preset: a fixed-prompt, two-tool coding-agent composition. The persona is the complete system prompt, so global identity, Web orientation, tool guidance
Creation Mode composition at b150a55
The cordis agent preset: the standard coding agent, plus the ability to read and write the runtime it is running in. It exists so a person can ask an agent to author another age