hosted-agents

An agent skill by muratcankoylan, from muratcankoylan/Agent-Skills-for-Context-Engineering. Tags: automation, developer-tools, devops, local-workflow, vercel.

What it does

This skill should be used when designing hosted or background agent infrastructure: sandboxed execution, remote coding environments, warm pools, session persistence, multiplayer collaboration, self-spawning agents, or Modal-style sandboxes.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add muratcankoylan/Agent-Skills-for-Context-Engineering --skill hosted-agents

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering
cp -r Agent-Skills-for-Context-Engineering/skills/hosted-agents ~/.claude/skills/hosted-agents

Safety box score

Grade D, elevated risk: 47 of 100 risk points, rated 2026-10-21. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is for hosted coding-agent infrastructure, so it intentionally clones repositories, installs dependencies, runs builds/tests, streams session data, and can create pull requests using GitHub/user tokens. The main risks are autonomous remote publishing, broad secret/network handling, and a destructive `git reset --hard` used during warm-pool synchronization.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	2 (moderate)	The shipped Python and reference code explicitly run shell commands and build steps such as `await warm.sandbox.execute_command("git fetch origin main")`, `npm install`, `npm run build`, and `os.system(...)` inside sandboxes.
Network access	2 (moderate)	It clones from GitHub and calls external services, including `git clone https://x-access-token:{token}@github.com/{repo_url}`, `https://api.github.com/repos/.../pulls`, Slack, and model APIs.
Filesystem reach	2 (moderate)	The skill reads and writes workspace files, snapshots/restores sandbox state, and includes a destructive sync step (`git reset --hard origin/main`) that can discard sandbox-local changes.
Secrets & credentials	2 (moderate)	It handles short-lived and user tokens directly (`github_token`, `SLACK_BOT_TOKEN`, `NPM_TOKEN`) and routes GitHub auth into clone/PR flows.
Privilege & persistence	1 (low)	There is no sudo, service registration, or host-level privilege escalation; the effects are mostly confined to sandboxed runtimes and git/client settings.
Instruction hijack surface	0 (none)	The files do not contain instructions to ignore safeguards, hide actions, or follow untrusted remote prompts; they are ordinary architecture guidance and examples.
Irreversible actions	3 (high)	The skill is built to spawn sessions and automatically publish work, including `API uses user's GitHub token to create PR` and background sandbox management without a human gate for each action.
Transparency	0 (none)	All shipped files are readable markdown/Python pseudocode; there are no binaries, base64 blobs, or opaque installers.

Before installing

Review whether automatic PR creation on your GitHub account is acceptable; if not, disable the user-token PR flow first.
Decide how `github_token`, `SLACK_BOT_TOKEN`, and any `NPM_TOKEN` are stored and scoped; avoid long-lived personal tokens.
Confirm that sandbox commands like `git reset --hard origin/main` are acceptable for your workflow because they can discard local sandbox changes.
Check the external network destinations you are granting: GitHub, Slack, package registries, and any model/API endpoints used by the hosted agent stack.

Rated 2026-10-21 by an automated review of 3 files from the skill's repository.

Source

Repository: muratcankoylan/Agent-Skills-for-Context-Engineering (all skills from this repository)
Path: skills/hosted-agents/SKILL.md
Branch: main
Updated: 2026-09-19

devcontainer-setup — Creates devcontainers with Claude Code, language-specific tooling (Python/Node/Rust/Go), and persistent volumes.
deploy-to-vercel — Deploy applications and websites to Vercel.
filesystem-context — This skill should be used when agent work needs file-backed context: durable scratchpads, tool-output offloading, just-in-time discovery.
harness-engineering — This skill should be used when designing autonomous agent harnesses: research loops, evaluation scaffolds, locked and editable surfaces, durable logs.
bambu-labs — Dry-run, upload, and cautiously initiate local Bambu Lab print jobs from validated plain `.gcode`, using Bambu LAN FTPS/MQTT handoffs.
cad — Create/edit parametric CAD models, organize CAD projects, export STEP/STL/3MF/GLB files, resolve prompt references, and measure geometry with cadgen.