SandBase Harness v0.3.2 and 88 installable Agent Skills #4
denial123789
announced in
Announcements
Replies: 1 comment
Update — third-party directory acceptance and clearer quick startsBoth projects are now listed in the independently maintained Awesome DeepSeek Harness directory (413 stars) after PR #141 was reviewed and merged. We also tightened the first-run experience:
|
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
We have been tightening the open-source path for builders who need both a reliable agent runtime and reusable, inspectable workflows.
SandBase Harness v0.3.2
SandBase Harness is a local-first TypeScript agent runtime for OpenAI, Anthropic, and OpenAI-compatible providers, including DeepSeek V4. It provides sandboxed sessions, MCP tools, approval boundaries, event streams, audit and replay, plus a native DeepSeek Harness bundle.
The immutable v0.3.2 release adds
Dockerfile.mcpfor a reproducible containerized MCP bridge. Its release gate passed typechecking, tests, production builds, package validation, and release smoke checks.SandBase Skills v0.3.2
SandBase Skills contains 88 installable Agent Skills for research, social intelligence, marketing, SEO, and business workflows across Codex, Claude Code, Cursor, Gemini CLI, and other compatible agents.
The flagship
multi-source-searchworkflow starts with search and page-reading capabilities already available to the host agent, so the first run needs no SandBase account or installation:Install it into Codex when you want to keep it:
It cross-checks independent sources, preserves credible disagreement, reports evidence gaps, and includes an offline validator for its claim-to-source ledger. SandBase MCP is optional provider expansion for Tavily, Exa, Scholar, and Cloudsway coverage.
Both projects are Apache-2.0 licensed. We especially value reproducible feedback from builders running local models, DeepSeek, MCP servers, or research-heavy agent workflows. If either project solves a problem for you, try it, open an issue, or star the repository so other builders can discover it.
All reactions