vallum 0.8.16

Security boundary between AI coding agents and your shell — redacts secrets, neutralizes prompt injection, sanitizes untrusted terminal output, audits every command.
Documentation
1
2
3
4
5
6
7
8
9
10
# pre-commit integration: requires a locally installed `vallum`
# (https://github.com/kahramanemir/Vallum#install). language: system is
# deliberate — building from source on every hook run is not acceptable.
- id: vallum-scan
  name: vallum scan
  description: "Static scan of agent skill/context files and MCP configs (secrets, injection, risky commands)."
  entry: vallum scan
  language: system
  pass_filenames: true
  files: '(^|/)(SKILL\.md|CLAUDE\.md|AGENTS\.md|GEMINI\.md|\.cursorrules|\.mcp\.json|mcp\.json|claude_desktop_config\.json)$|\.mdc$'