Skip to content

Security: stylusnexus/defect-scan

SECURITY.md

Security Policy

Reporting a vulnerability

Please do not open a public issue for security problems.

Report privately via GitHub's private vulnerability reporting: go to the repository's Security tab → Report a vulnerability (or the Advisories section). This keeps the report confidential between you and the maintainers until a fix ships.

When reporting, include where you can: affected file/subcommand, a reproduction or proof-of-concept, the impact, and any suggested remediation.

We aim to acknowledge a report within a few business days and will keep you updated through the advisory.

Scope — what we care about most

defect-scan runs analyzers and shells out to git/gh against code it scans, and can file issues. The threat model that matters most here:

  • Command/argument injection from scanned content (the plugin's own pattern P4). Findings and file paths flow into gh and shell helpers — they must never be able to inject flags or execute commands. See skills/scan/lib/detect.sh (argv is quoted; issue bodies are passed via file, never argv; no eval).
  • Origin-gated tool execution. A scanned repo's user/project profile must never cause its declared analyzers to auto-run — built-in profiles auto-run, user/project profiles are confirmed first (SKILL.md Stage 2). A bypass of this gate is a vulnerability.
  • Unintended writes. --fix and --file-issues are the only write paths; both are opt-in and gated. A scan that mutates files or the tracker without the corresponding flag + confirmation is a vulnerability.

Supported versions

This is an actively developed plugin; security fixes target the latest released version on the default branch.

There aren't any published security advisories