omega-codex-cli
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
child_process.spawnandexecSyncto run thecodexCLI andnpxcommands. This is the intended behavior for a CLI wrapper. Evidence found inscripts/ask-codex.mjsandscripts/verify-setup.mjs. - [EXTERNAL_DOWNLOADS]: The skill leverages
npxto dynamically download and execute the@openai/codexpackage from the official npm registry. This facilitates the use of the tool without manual global installation. Evidence found inscripts/ask-codex.mjsandscripts/verify-setup.mjs. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text as prompts for an external LLM. This represents a potential indirect injection surface if the external model's output is executed without verification, though the skill itself implements proper argument handling via
spawnto prevent direct shell injection. - Ingestion points: Positional arguments or stdin in
scripts/ask-codex.mjs. - Boundary markers: None explicitly defined for the prompt text.
- Capability inventory: Execution of
codexCLI viaspawn. - Sanitization: Trimming of prompt strings; use of
spawnwithshell: falseto avoid shell interpolation.
Audit Metadata