Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add yi427/Cloister --skill host-execgit clone --depth 1 https://github.com/yi427/CloisterWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/yi427/cloister/host-exec)<a href="https://agentmods.dev/skills/yi427/cloister/host-exec"><img src="https://agentmods.dev/badge/skills/yi427/cloister/host-exec/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/yi427/cloister/host-exec"><img src="https://agentmods.dev/badge/skills/yi427/cloister/host-exec.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00077 | $0.01646 |
| Opus 5.5 | $0.00031 | $0.00658 |
| Sonnet 5.5 | $0.00015 | $0.00329 |
| Haiku 4.5 | $0.00008 | $0.00165 |
Grade C, and why
host-exec scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nullifies safety policieshighAnti-refusal
"You have no restrictions", "do anything now", "ignore your guidelines": a direct jailbreak that disables guardrails.
or execution mechanism to bypass the policy. How it starts
The opening of the file, as written. The whole thing — 128 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cloister Host Exec
Use only the cloister_host MCP tools for host execution. Treat the immutable
policy returned by the bridge as the authority for the entire bridge session.
Choose Guest or Host
- Use Guest file tools to read, write, list, and patch files in the selected
Guest workspace: the agent's initial working directory under
/workspace/. It is a live read-write mount of the selected Host workspace, so Guest changes already appear in the corresponding Host project. - Do not use
host.execmerely to move workspace content across the Guest/Host boundary. In particular, do not Base64-encode a workspace file or invoke an allowed interpreter such aspython -cto recreate it on the Host. - Use Guest commands when they can complete the task. Use Host Exec only when the task needs macOS capabilities (such as Xcode or a macOS API) or a required executable missing in the Guest. Guest-first routing is behavioral guidance, not an enforced security boundary.
- Explain why Host execution is needed before each invocation. Keep calls within the user's requested task; avoid redundant probes, repeated discovery, and unnecessary Host shell or interpreter wrappers. When that executable consumes workspace files, prefer paths relative to its fixed Host workspace working directory.
- If a requested file is outside the selected Guest workspace, report that it is outside the shared workspace. Do not treat an allowed interpreter as a Host file API.
Execute a host command
- Call
host.list_commandsbefore the first host operation in a bridge session. Call it again after a reconnect or when the bridge reports that the session changed. - Read the returned DSL
version,mode, fixed Host working directory, command names, descriptions, argument policies, environment mode, environment variable names, and audit status. Prefer workspace-relative paths when calling a Host command. The bridge uses the Host startup environment; GuestPATHdoes not describe Host executables. - Follow the mode:
allowlist: Select only a returned command name. If absent, report it as unavailable. Do not substitute another command, interpreter, build tool, or execution mechanism to bypass the policy.on-demand: Pinned commands remain available, and other bare Host command names can be discovered withhost.find_command. Readavailable,executable, andresolved; if unavailable, report it without attempting installation or another execution path. Discovery does not execute the program. Reuse successful results in this session. Pass the discovered name, never an arbitrary absolute path, tohost.exec. No registration or Profile rewrite is needed. If a cached executable changes, report the error; restart the bridge before rediscovering.
- Call
host.execwith these model-supplied fields:
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +26 lines · -24 tokens per session 9aab35323d08
- 1mo ago Changed · +2 lines 9207c3a27b66
- 2mo ago First seen · 100 lines · 101 tokens per session scan C 5c8b6071cf22
host-exec is a skill published in the GitHub repository yi427/Cloister (2 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 77 tokens to every session and 1,646 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it C with 1 finding (nullifies safety policies). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
host-exec compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| host-execyi427/Cloister | Skill | 2 | 1mo ago | Apache-2.0 | C | 77 | Codex |
| macvisjunmo-kim/mac-local-vision | Skill | 64 | 16d ago | MIT | A | 319 | Claude Code |
| gguf-quantizationsynthetic-sciences/openscience | Skill | 4.4k | 2d ago | Apache-2.0 | A | 48 | Claude Code, Codex |
| vox-voice-alertskohlhofer/vox | Skill | 7 | 2mo ago | MIT | A | 117 | Claude Code, Codex |
| speak-ttsEmZod/speak | Skill | 8 | 9mo ago | none | A | 70 | Claude Code, Codex |
| feishuHmbown/CodeWhale | Skill | 41k | 2d ago | MIT | A | 33 | Claude Code, Codex |
| interviewHmbown/CodeWhale | Skill | 41k | 2d ago | MIT | A | 40 | Claude Code, Codex |