dsh-permission-rules
PerryLink
Declarative allow/deny/ask permission rules on the tools/pre-execute waterfall: tool/argument/path/agent matching, session-log audit, dry-run mode, hot reload.
PerryLink2487Updated 3d ago
Second-model AI auto-review on the approval answerer chain: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default.
Practical score
83 / 100
Setup
Needs setup
License
Apache-2.0
Language
TypeScript
Copy the command, or hand the whole task to your harness agent.
git clone https://github.com/PerryLink/dsh-auto-review.gitPlease install the DeepSeek Harness plugin "dsh-auto-review" for me. - GitHub repository: PerryLink/dsh-auto-review (https://github.com/PerryLink/dsh-auto-review) - Type: a Cordis plugin (installed into a DSH profile) - The plugin declares a DSH version requir…
A cleaned excerpt from the project's own documentation.
dsh-auto-review 让第二个模型在审批链上替你裁决——人类零点击,危险操作零漏网。 面向 DeepSeek Harness 的第二模型 AI 审批插件:当动作越过沙箱边界,一个只读审查子代理读取证据并返回带理由的裁决,默认失败关闭。 安装 · 配置 · FAQ · 简体中文 · English · Español · Português · हिन्दी [![CI](https://img.shields.io/github/actions/workflow/status/PerryLink/dsh-auto-review/ci.yml?branch=main&label=CI&…
Open on GitHubSimilar tools from the same corner of the ecosystem.
PerryLink
Declarative allow/deny/ask permission rules on the tools/pre-execute waterfall: tool/argument/path/agent matching, session-log audit, dry-run mode, hot reload.
PerryLink
Bounded, layered, approval-gated, auditable cross-session memory: typed ctx.memory seam, zero-dependency SQLite provider, memory tool and frozen snapshot injection, plus the dsh-memory-protocol v1 rehearsal with an adapter registry and a conformance suite.
PerryLink
Prompt-injection, jailbreak, and secret-leak detection on the official seams: an Aho-Corasick engine over rules ported from Prompt-Injection-Payloads, Jailbreak-Detector, and Secret-Key-Leaker-Detect gates user messages, tool arguments, and tool results with allow/ask/block tiers, sanitized defend/detection audit events, a defend_report tool, and a destructive-delete command guard.