Skip to content

Security: GrubbyLee/skill-manager

Security

SECURITY.md

Security Policy

skill-manager inspects local skill packages and MCP metadata across supported AIDE tools. Most commands are read-only for AIDE data, while explicit lifecycle commands can stage, install, update, back up, or restore skill directories. Security reports are welcome for either surface.

Supported Versions

The main branch and latest GitHub Release are supported.

Reporting a Vulnerability

Please open a private security advisory or contact the maintainer through GitHub:

https://github.com/GrubbyLee/skill-manager/security/advisories/new

If the issue is not sensitive, you can open a regular issue:

https://github.com/GrubbyLee/skill-manager/issues

Sensitive Data

Do not include API keys, passwords, cookies, tokens, private keys, full session logs, or private filesystem paths in public issues or discussions.

For MCP-related reports, describe the transport and command shape, but do not paste environment variable values.

Project Boundaries

  • Inventory, reporting, recommendation, graph, audit, outdated, and lifecycle verification commands are designed to be read-only for AIDE data, though they may refresh skm-owned files under ~/.skill-manager.
  • Explicit setup, install, update, rollback, profile apply, state set, sessions --clean, disable, and enable commands can modify supported AIDE files.
  • Skill install/update must preserve staging validation, complete-package static audit, the high-risk policy gate, and confirmation. Update/rollback must preserve instance-scoped backups and atomic replacement with restoration on failure.
  • Repository packages are copied but never executed by skm. MCP env values and secrets are outside the intended collection surface.

There aren't any published security advisories