build-mcpb
What this skill does
Build and distribute local MCP servers (packaged with their runtime) that can interact with the local filesystem, desktop apps, or OS, without requiring Node/Python on the user's machine.
github/anthropics - Privilege Escalation - 33.4k stars
Threat analysis
Skill info
pkg:github/anthropics/claude-plugins-official@920824c?skill=build-mcpbAssessments (2)
Privilege Escalation
Privilege Escalation via local-llm-review
references/local-security.md
MCPB provides no sandbox. There's no `permissions` block in the manifest, no filesystem scoping, no network allowlist enforced by the platform. The server process runs with the user's full privileges Security Misconfiguration
Security Misconfiguration via local-llm-review
references/local-security.md
Tool inputs are untrusted, even though they come from an AI the user trusts. A prompt-injected web page can make Claude call your `delete_file` tool with a path you didn't intend.Badge
Add the Anomity scan badge for build-mcpb to your README.
How Anomity governs this at runtime
Scan-time vetting tells you what a skill says it will do. Anomity's Endpoint Sensor sees what agents actually do: it discovers skills alongside every other AI artifact on the endpoint, and runtime governance can allow, deny, or log the tool calls a skill triggers. Policy violations route to your SIEM, Slack, email, or Jira, backed by a queryable 90-day audit trail.
Book a 30-minute demo to see your own skill inventory.
Methodology and disputes
Every skill is assessed by the Anomity Skill Intelligence engine against its public source; findings indicate risk patterns, not confirmed exploitation. Maintainer of build-mcpb? Report an issue or request a rescan.
