@jkrandom-sudo
Security audit for DeepSeek Harness plugins: static permission profiling with file/line evidence plus a runtime sentinel that gates credential access and unknown-host egress, requiring approval before risky actions.
@zuoguyoupan2023
Highlights parts of speech in rendered Markdown within DeepSeek Harness Web UI, coloring nouns, verbs, adjectives, and more.
@JMOKSZ
A Feishu (Lark) entry point for DeepSeek Harness, allowing users to drive DSH agents from a Feishu bot with streaming cards, ask/approval buttons, and attachments.
@Jolly-J
A DSH WebUI plugin that displays DeepSeek account balance and per-session cost estimates in a sidebar card, with auto-refresh and detailed token breakdowns.
@bobcat848
Calculate the real-time cost of DeepSeek API calls made by DeepSeek Harness.
@YiHarvest
Local-first failure evidence capsules for DeepSeek Harness: on tool/agent failure, automatically package sanitized context (timeline, git state, environment, plugin list) into a ZIP for debugging.
@SPYQWER1
Registers three model tools backed by the ChatGPT Codex transport: web search, image generation, and image vision. Uses existing Codex login, no npm dependencies.
@moon09300731
Automatic approval gate for DeepSeek Harness: Flash model pre-judges sandbox escapes, auto-approves safe operations, routes hard risks to human approval with fail-safe, learning from confirmed operations.
@Khellendros97
Adds subscription-based OAuth login support to DSH for model providers including ChatGPT, Claude, Grok, and Kimi. Allows logging in with subscription accounts instead of API keys. Implements OAuth flows (authorization code with PKCE, device flow) and routes inference requests to respective APIs.
