DeepSeek Harness Plugins · Memory
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Offline deterministic verification of body-free DSH tool-result custody receipts, checking digests, byte budgets, and projection mappings without reading result bodies or spill artifacts.
Automatically patches hardcoded English labels in the DSH Web UI to Chinese on every startup, surviving DSH upgrades without manual intervention.
Auto snapshot guard for DSH: saves config and plugin snapshots on exit, verifies on startup, supports one-click rollback, single-zip export, and isolated recovery.
An intelligent memory plugin for DSH that tracks user preferences, tool usage, and project context to deliver personalized model and tool recommendations, with local persistent storage and a web visualization viewer.
Adds an Archive Manager tab to Settings, listing archived sessions and enabling true disk deletion of session logs, archive markers, and projection cache.
A gated, file-based long-term memory plugin that only records verified facts and rules, with a session notepad and settlement protocol, requiring no vector DB or cloud.
A unified context console for DeepSeek Harness covering trajectory inspection, Prompt/Skill/MCP/Tools management, cache observation, message forging, and sessionlog repair.
Dual-ledger cross-session memory for DSH that auto-extracts project facts and progress from past sessions and injects them into every new conversation's system context.
A minimal snippet toolbox that stores high-frequency commands, code, and prompts in a local JSONL file, with zero database, zero network, and on-demand retrieval via tool calls.
Read-only health check for your dsh environment covering disk, memory, network, and directory bloat, producing tiered reports with historical trend comparison.
A three-tier memory foundation for agents with L1-L3 layered memory, gate approval, event logging, and a TDD loop, supporting both Claude Code and DeepSeek Harness.
A high-density session summarization skill for coding agents, supporting progress review, context compaction, and cross-task handoff with evidence-driven prioritized recommendations.