Task fit
30%15 points for catalog classification, 8 for an explicit category tag, and 7 when the name, description, or tags match: security, guard, credential, approval, audit, sandbox, redaction, safe, permission.
Security, guardrail, audit, credential, and safety projects ranked by public ecosystem evidence—not assumed trust.
Published selection rule
Standalone, installable public repositories classified as Security. Upstream, directory/list, archived, adaptation-required, watch, removed, placeholder, and non-plugin records are excluded. This is discovery, not a source-code security audit.
Ties are resolved by maintenance score, then Stars, then repository name.
100
point scale
15 points for catalog classification, 8 for an explicit category tag, and 7 when the name, description, or tags match: security, guard, credential, approval, audit, sandbox, redaction, safe, permission.
25 points if updated within 30 days; then 20 within 90, 14 within 180, 8 within 365, and 2 thereafter.
Stars are bucketed, not directly sorted: 20 at 1,000+, 17 at 250+, 14 at 50+, 10 at 10+, 7 at 3+, and 4 at 1+.
Forks are bucketed: 10 at 100+, 8 at 25+, 6 at 10+, 4 at 3+, and 2 at 1+.
8 points for Awesome + Topic, 6 for Awesome only, or 2 for Topic only; 4 for an 80+ character English description or 2 for 30+; 3 for 3+ tags or 1 for 1–2.
Scores use public catalog and repository metadata. They do not replace hands-on evaluation, permission review, or a security audit.
Scored results
sandbaseai / sandbase-harness
Open-source CMA-compatible agent runtime for any model, with MCP tools, sandboxed sessions, audit, replay, and a local console. Includes a native DeepSeek Harness bundle over stdio MCP.
Why it placed here
cofy-x / axern
Open-source sandboxes for AI agents, untrusted code execution, and durable services.
Why it placed here
pulseaiclub / phi
a coding Agent from pi. ∞ providers, sub-agents, hashline edits, and a permission gate
Why it placed here
openguardrails / openguardrails
The vendor-neutral protocol for AI agent safety & security — and the neutral benchmark that ranks the vendors.
Why it placed here
Anionex / dsh-turn-rewind
deepseek harness对话和代码状态回退插件 | DSH — rewind conversation and workspace state, powered by a persistent Change Ledger
Why it placed here
Anionex / dsh-computer-use
为 DeepSeek Harness 提供电脑控制插件:新鲜 Accessibility 观测、过期状态拒绝、作用域权限与安全输入(目前支持macos)|Accessibility-first macOS Computer Use bundle for DSH with fresh observations, stale-state rejection, scoped permissions, and safe input.
Why it placed here
Zhenyu98 / dsh-context-doctor
DSH 上下文注入审计插件:统计 AGENTS.md 指令链/技能目录/工具 schema 的 token 成本,检测重复与冲突;Web UI 圆环面板 + context_audit 工具。Context Doctor for DeepSeek Harness: audit instruction-chain / skill catalog / tool schemas token cost.
Why it placed here
omdsh-dev / dsh-custom-tool
Create and manage sandboxed JavaScript tools for DeepSeek Harness with a Monaco editor and model-driven tool lifecycle.
Why it placed here
omdsh-dev / dsh-security-audit
DSH 本机安全审计插件:配置/插件来源/会话/网络暴露面,只读脱敏风险报告
Why it placed here
N0zoM1z0 / vocaloid-mcp
An agent-native MCP for composing, tuning, rendering, mixing, and auditing native VOCALOID3/4 projects — built just for fun.
Why it placed here