codex-setup

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands using Bash and Node.js tools to automate the installation of Git hooks, copy utility scripts, and create project directories.
  • [PERSISTENCE_MECHANISMS]: Automates the installation of Git hooks (commit-msg and pre-push) into the local repository. These hooks act as persistent execution points that run scripts during the development lifecycle.
  • [INDIRECT_PROMPT_INJECTION]: Ingests metadata from package.json and CLAUDE.md to generate AGENTS.md, which contains instructions that guide agent behavior. This creates a surface where manipulated project files could indirectly inject malicious instructions into the agent's context.
  • Ingestion points: Reads from package.json and CLAUDE.md in the project root.
  • Boundary markers: None are explicitly defined for the data extracted from these files.
  • Capability inventory: Access to shell execution (bash, node), file system modifications, and Git toolchain.
  • Sanitization: Implements output validation, including a 24 KiB size limit and verification for unresolved placeholders.
  • [SAFE]: Implements defensive shell scripting in the installed Git hooks, utilizing absolute paths for binaries and bash -p to prevent function injection and unintended environment variable loading.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 01:44 PM
Security Audit — agent-trust-hub — codex-setup