hannnnnnnny

GuardRun — Security skill for Claude Code

Security community

GuardRun: a safety-first AI agent harness for DeepSeek and Claude, with a path sandbox, shell command policy, approvals, secret redaction and an audit log.

How to install GuardRun

This entry records only its repository, not the path inside it, so there is no exact command to give. Open hannnnnnnny/GuardRun and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What GuardRun does

GuardRun: a safety-first AI agent harness for DeepSeek and Claude, with a path sandbox, shell command policy, approvals, secret redaction and an audit log.

Alternatives in Security

Security Guardian — Comprehensive security analysis for RTK CLI tool, focusing on command injection, shell escaping, hook security 11.9k ★
ThinkWatch — Self-hosted AI API and MCP gateway for organizations: SSO and RBAC, per-user identity for MCP tool calls, PII 817 ★
OpenTag — Open-source, channel-native agent gateway for Slack 499 ★

README

GuardRun

[![CI](https://github.com/hannnnnnnny/GuardRun/actions/workflows/ci.yml/badge.svg)](https://github.com/hannnnnnnny/GuardRun/actions/workflows/ci.yml) [![CodeQL](https://github.com/hannnnnnnny/GuardRun/actions/workflows/codeql.yml/badge.svg)](https://github.com/hannnnnnnny/GuardRun/actions/workflows/codeql.yml)

A powerful, safety-first agent harness for DeepSeek and Claude models, written in Python.

GuardRun gives Claude real tools (files, search, and a shell) inside a single workspace directory, and puts a layered safety system between the model and your machine: a path sandbox, a shell command policy, approval modes, secret redaction, prompt-injection flagging, budgets, and an audit log.

you ──task──▶ Agent loop ──request──▶ Claude
                 ▲   │
        results  │   ▼ tool calls
                 │  ToolExecutor ──▶ validate ─▶ SafetyGate ─▶ tool ─▶ redact / flag ─▶ result
                 │                                   │
                 └──────── events ──▶ audit log, console

Install

pip install -e ".[dev]"

Python 3.10 or newer. GuardRun uses **DeepSeek** by default (`deepseek-v4-pro` via DeepSeek's Anthropic-compatible API). Put your key in an environment variable; GuardRun reads it from there and never stores it:

export DEEPSEEK_API_KEY=...        # PowerShell: $env:DEEPSEEK_API_KEY = "..."
guardrun doctor                    # confirms the key is visible (without printing it)

To use Claude instead, set `provider = "anthropic"` under `[model]` in `guardrun.toml` (or pass `--provider anthropic`) and provide `ANTHROPIC_API_KEY` or an `ant auth login` profile.

Quick start

guardrun init                                  # optional: write guardrun.toml
guardrun run "add type hints to utils.py and run the tests"
guardrun chat                                  # multi-turn session
guardrun run "continue" --resume 3f2a9c1b7d4e  # pick a session back up
guardrun usage