Skip to content

Security: produtoramaxvision/maxvision

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Do not open a public GitHub issue for security concerns.

If you discover a vulnerability in maxvision:

  1. Email producoes.maxvision@gmail.com with subject [SECURITY] maxvision: <short description>.
  2. Include:
    • A description of the vulnerability
    • Steps to reproduce (if possible)
    • The version of the plugin (.claude-plugin/plugin.json version field) and the commit sha you tested against
    • Any proposed fix or mitigation

We aim to acknowledge reports within 5 business days and to release a fix within 30 days for confirmed high/critical issues.

Threat model

This plugin orchestrates third-party skill installation. The primary risks are:

Risk Mitigation
Malicious upstream skill cloned into ~/.claude/skills/ Single confirmation per install; tier_4 entries always show warning text; only public, license-checked sources in the catalog
Compromised marketplace plugin Plugin install commands are interactive (/plugin install requires user confirmation in Claude Code's UI) — no programmatic install
Stale .maxvision-source.json claiming false provenance check-version always re-fetches upstream sha when cache is stale or --force-check passed
Cache poisoning of ~/.claude/cache/maxvision/version-check.json Cache is read/write by the user account that owns Claude Code; if the user account is compromised, this plugin is not the weakest link
Unauthenticated gh api rate limiting Plugin requires gh auth login and surfaces the 60/h vs 5,000/h difference clearly
Force-push or destructive git operation Plugin only uses git pull --ff-only and sparse-checkout; never force-pushes or resets

Out of scope

  • Vulnerabilities in third-party skills installed via the catalog. Report those to the respective upstream repos.
  • General Claude Code platform vulnerabilities. Report to github.com/anthropics/claude-code.
  • Vulnerabilities in gh CLI, git, jq, or npx. Report to upstream maintainers.

Disclosure

Confirmed vulnerabilities will be disclosed in CHANGELOG.md after a fix is released, with credit to the reporter (unless the reporter requests otherwise).

There aren't any published security advisories