Skip to content

Security: yannmenec/inspectrum

SECURITY.md

Security Policy

Supported versions

Version Supported
0.2.x ✅ active
< 0.2 ❌ not maintained

Reporting a vulnerability

Please do not open public GitHub issues for security reports.

Report privately via GitHub Security Advisory: https://github.com/yannmenec/inspectrum/security/advisories/new

You'll receive an acknowledgement within 5 business days. After triage, we'll work with you on a fix and a coordinated disclosure timeline.

In scope

Reports that materially affect users of inspectrum itself:

  • Subprocess argv injection or shell-escape in the reviewer CLI wrappers (src/reviewers/).
  • Path traversal in MCP resource handlers (inspectrum://sessions/...).
  • Secret-leak risks introduced by inspectrum's own code (e.g. unintended logging of API keys, world-readable session files, etc.).
  • Unsafe handling of plan / context input that could trigger arbitrary file read/write outside ~/.inspectrum/.

Out of scope

  • Vulnerabilities in upstream reviewer CLIs themselves (Claude Code, Codex, Gemini, Kimi, Qwen, Ollama). Please report those to their respective maintainers.
  • Issues that require an attacker to already have local code execution as the user running inspectrum.
  • Bugs in third-party MCP hosts (Claude Code Desktop, Cursor, etc.).
  • Network-level attacks against openrouter.ai or any other upstream provider — those are the provider's responsibility.
  • Anything in _decisions/, AGENTS.md, or test fixtures (these are not shipped in the npm tarball).

Disclosure

Once a fix is released, the advisory will be made public on the same GitHub Security Advisory page. Credit to reporters is given by default unless you ask otherwise.

There aren't any published security advisories