Hosted Agent Infrastructure

Trust Score 79/100

Design patterns and operational guidance for running agents in remote sandboxed environments with warm pools, snapshots, and session orchestration.

triggers:hosted agentsandboxwarm poolsnapshotsessionbackground agentimage build

GitHub SKILL.md

What it does

Describes architecture and operational patterns for running AI agents in hosted sandbox environments. Covers image build strategies, warm-pool management, snapshot/restore, auth flows, and techniques to minimize cold-start latency.

When to use it

Activate this skill when building background agent infrastructure, pre-warmed sandboxes, or systems where agents spawn sub-sessions and create PRs or push code on behalf of users. Use it for decisions about scaling, session isolation, and real-time streaming.

What's included

Scripts: present in the repository for image/sandbox tooling (see scripts/ in repo)
References: architectural notes and patterns under references/
Instructions: practical checklists for sandbox lifecycle, image rebuild cadence, warm-pool sizing, authentication, snapshot policies, and client integrations.

Compatible agents

Agent platforms that integrate with sandboxed runtimes and GitHub (e.g., Claude Code, Cursor, Copilot-style tooling, and custom server-first agents).

Audit Summary

Well-structured reference skill for designing hosted agent infrastructure with sandboxed execution, warm pools, and session orchestration. SKILL.md is thorough and well-organized. The bundled sandbox_manager.py script is pseudocode with architectural patterns — it runs but crashes in the demo because _cold_start and other infrastructure-specific methods are stubs (pass). DeprecationWarning on utcnow(). No security concerns beyond using f-strings for git config (minor shell injection surface in sandbox context, mitigated by sandbox isolation).

Watch Out

Script demo crashes due to NoneType from _cold_start stub — not production code, just architectural pseudocode
Uses deprecated datetime.utcnow() — should use datetime.now(datetime.UTC)
Git config commands use f-strings with user.name/email — shell injection risk if untrusted input, though mitigated by sandbox isolation
GitHub token embedded in clone URL via f-string — standard pattern but be aware of logging exposure

Notes

This is a reference/architectural skill, not a runnable tool. The script is well-designed pseudocode demonstrating patterns. SKILL.md is exemplary — clear scope, boundary decisions, integration map, practical checklist, and gotchas. Docked code quality for stubs that crash in demo and deprecated datetime usage. Docked security slightly for f-string command construction. Architecture score is high due to excellent separation of concerns, clear frontmatter, references section, and progressive disclosure.

Information

Repository: open-agent-hub
Stars: 886

Trust Score

Overall79

Security80

Code Quality72

Architecture85

Usefulness70

Related Skills

Development Worktree

Create an isolated git worktree for feature work, auto-run project setup, and verify a clean test baseline before development.

Readwise Reader Document Management

Manage Readwise Reader documents: list, save, search, move, tag, highlight, export and bulk-edit via official and custom CLIs.

Bounty Hunter — Atlas

Persona skill: 'Atlas' — a profit-focused developer persona for discovering, evaluating and executing paid bounties or freelance tasks with ROI-aware workflows.

Junshi — Research Advisor

Daily strategic research advisor that scans arXiv/venues, digests papers, and proposes bold, ranked research ideas tailored to the user's profile.

Full Stack Builder

End-to-end builder that scaffolds, implements, tests, and optionally deploys web and API applications from a natural-language specification.

ezBookkeeping API Tools

Command-line API tools for ezBookkeeping: record and query transactions, retrieve accounts/categories/tags, and fetch exchange rates for self-hosted personal fi

Feishu Voice Sender

Convert MP3s and send them as native Feishu voice messages (playable voice clips) to users or groups.

Claw Bench

Benchmarking skill that guides an agent through a structured suite of capability tests and reporting steps for leaderboard submission.

Back to Skills