Skip to content
Development
Skill

/select-analysis-isolation

Select isolation before inspecting or executing untrusted content. Use for local analysis, a container, Linux or macOS VM, remote sandbox, or spare device with defined network, mounts, credentials, snapshots, evidence export, and teardown.

From plugin
socket
7200 skills5 MCP
Install
$ npx -y skills add gaelic-ghost/socket --skill select-analysis-isolation --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/select-analysis-isolation

Context preview

The summary Claude sees to decide when to auto-load this skill.

Select isolation before inspecting or executing untrusted content. Use for local analysis, a container, Linux or macOS VM, remote sandbox, or spare device with defined network, mounts, credentials, snapshots, evidence export, and teardown.

SKILL.md

select-analysis-isolation.SKILL.md
name: select-analysis-isolation
description: Select isolation before inspecting or executing untrusted content. Use for local analysis, a container, Linux or macOS VM, remote sandbox, or spare device with defined network, mounts, credentials, snapshots, evidence export, and teardown.

Select Analysis Isolation

Overview

Choose the smallest environment that contains the behaviors the analysis may trigger and can still reproduce the target platform. Treat isolation as a set of controls to verify, not a label supplied by a product.

Read [references/isolation-matrix.md](references/isolation-matrix.md) before selecting an environment for execution or privileged tooling.

Workflow

1. Identify the behavior to contain.

  • Record target OS/runtime, expected privilege, kernel or device access, persistence, networking, anti-VM behavior, and data sensitivity.

2. Choose the environment.

  • Use local read-only inspection for non-executing metadata and text extraction.
  • Use a disposable container for bounded Linux user-space tooling when VM-level isolation and target-platform behavior are unnecessary.
  • Use a disposable VM for installers, services, dynamic behavior, or full-OS effects.
  • Use a macOS VM or spare Mac for macOS payload behavior; do not substitute a Linux container.
  • Use a remote sandbox only after data-egress approval and provider-capability review.

3. Remove ambient authority.

  • Exclude personal accounts, host directories, shared clipboard, drag/drop, host sockets, USB/device passthrough, SSH agent, developer certificates, browser profiles, password stores, cloud tokens, and unrelated secrets.

4. Constrain networking.

  • Default to no network or a simulated service.
  • If external traffic is required, allow only recorded destinations through a monitored boundary and capture DNS/routes/packets proportionately.

5. Establish lifecycle evidence.

  • Record base image or restore image, OS build, tool versions, configuration, snapshot, clocks, and baseline state.
  • Define exactly which evidence may be exported and how it will be scanned before reaching the host.

6. Verify teardown.

  • Stop the workload, export intended evidence, revert or destroy disposable state, revoke temporary credentials, and confirm no host share or forwarded port remains.

7. Hand an approved execution boundary to `prepare-isolated-analysis-lab`.

  • Use `apple-dev-skills:choose-macos-virtualization-shape` when the development boundary is still undecided.
  • Use `apple-dev-skills:virtualization-framework-workflow` when a custom macOS or Linux VM host must be implemented or diagnosed.
  • Use `reverse-engineering-skills:research-macos-security-control` when the lab must preserve an exact-build public/private control comparison and technical-note contract.
  • Treat a SIP-enabled macOS VM as the stable high-fidelity path for SIP-sensitive behavior; local sandbox, TCC, and failure injection are explicitly lower-fidelity approximations.

Stop Conditions

Stop before execution when the environment cannot reproduce the target platform, the isolation controls cannot be verified, or the task requires host secrets or privileges beyond the approved analysis plan. Selection alone does not authorize execution; require the prepared-lab record first.

Read more
Ships withsocket

Stuff for Agents on macOS Promo audio: Socket Codex Marketplace Promo

Get the whole plugin

Other skills on socket.