Installer tirith avec Homebrew, Nix

Liens du projet

• Page d'accueil
• Dépôt
• Page du gestionnaire de paquets

Commandes et alias

• tirith

Historique du projet

The tirith GitHub repository was created in February 2026. Its README and llms.txt describe an open-source, offline, no-telemetry tool that intercepts commands and pastes for homograph URLs, pipe-to-shell patterns, ANSI injection, credential leaks, data exfiltration, malicious AI skills/configs, and related threats.

By the 0.3.x releases in spring and summer 2026, the project had expanded beyond shell hooks into threat databases, MCP tools for AI coding agents, an LSP server, AI-config drift detection, custom rules, policy validation, CI/SARIF output, and package-risk scoring. The changelog shows rapid iteration around Bash enforcement, threat-intelligence updates, agent governance, prompt-injection detection, and installer verification.

Historique d'adoption

tirith is too new to have long historical adoption, but official docs already target package-manager installation and AI-agent workflows. The official quickstart lists Homebrew, cargo, and npm installation paths; the input package metadata lists Homebrew and Nix packages.

Its adoption niche is specifically modern package and agent hygiene: protecting curl|bash installs, homograph package URLs, AI config files, MCP server definitions, shell hooks, and CI scan outputs. That makes it a package-nerd tool because its value appears exactly at the moment developers paste install commands, run package-manager commands, or let coding agents execute shell operations.

Modes d'utilisation

Users install tirith, activate a shell hook such as eval "$(tirith init --shell zsh)", and then let the hook inspect commands before they run. Clean commands pass silently; suspicious commands can warn or block depending on shell and policy.

The CLI also supports explicit checks and scans: tirith check analyzes a command without running it, tirith scan checks files and directories, tirith run is a safer replacement for curl|bash style installers, tirith doctor reports protection status, and the MCP/LSP modes integrate with AI coding agents and editors.

Pourquoi les passionnés de paquets s'y intéressent

tirith is significant to package nerds because it is explicitly about the dangerous edges of package culture: pipe-to-shell installers, typosquatted package names, malicious packages, insecure package-manager flags, untrusted Docker registries, and AI-agent tool execution. Its short history should be treated as current-tool adoption, not legacy adoption.

Chronologie

• 2026-02-02: The tirith GitHub repository was created.
• 2026-04-21: tirith 0.3.0 added major shell-hook, threat-intelligence, policy, and CI features.
• 2026-06-16: tirith 0.3.2 expanded platform reach, agent governance, LSP/editor support, and operational-context checks.
• 2026-06-19: tirith 0.3.3 added evasion-resistant prompt-injection detection and output-data-exfiltration handling.

Related projects

• MCP-aware AI coding tools such as Claude Code, Codex, Cursor, Gemini CLI, VS Code, and Windsurf are named integration targets in official tirith docs.
• OpenSSF malicious package feeds, Datadog Security Labs datasets, abuse.ch feeds, OSV.dev, deps.dev, and Google Safe Browsing are listed by tirith as threat-intelligence or enrichment sources.
• Shells such as Bash, Zsh, Fish, and PowerShell are part of the supported hook surface.

Sources

• https://api.github.com/repos/sheeki03/tirith
• https://github.com/sheeki03/tirith#readme
• https://raw.githubusercontent.com/sheeki03/tirith/main/CHANGELOG.md
• https://tirith.sh/llms.txt
• source_facts.package-manager